WEBVTT
00:00:00.000 --> 00:00:06.799
Hello everybody, welcome to the fourth episode of our Talking Data and AI webinar series.
00:00:06.959 --> 00:00:10.800
My name is Lyle Richie, and I'll be your host today.
00:00:11.279 --> 00:00:15.199
Today we'll explore evolving skills and the data driven world.
00:00:16.079 --> 00:00:19.280
I'm delighted to introduce today's guest, Elizabeth Hollinger.
00:00:19.519 --> 00:00:29.760
Elizabeth has over 20 years' experience in data and how data has budget roles across a range of companies, including Deloitte, Avanti, and Graphics.
00:00:30.079 --> 00:00:34.880
So you could say to note a thing or two about reading high performance data and tech teams.
00:00:35.439 --> 00:00:41.280
And I'll just talk about that in more detail and about how to build adaptability as a course culture.
00:00:42.560 --> 00:00:45.920
So before I hand over to Elizabeth, just a couple of quick points.
00:00:46.640 --> 00:00:54.079
Elizabeth will talk to all today for around 30 minutes, and then there will be 50 minutes for a QA.
00:00:54.479 --> 00:01:04.799
So if you could please use the QA box and the chat function, and if you could thumbs up the questions you would like the answers to, and we will get those answered for you.
00:01:05.120 --> 00:01:09.439
So really looking forward to this, and I will hand over to Elizabeth.
00:01:09.680 --> 00:01:10.879
Elizabeth, over to you.
00:01:11.599 --> 00:01:15.200
Thanks very much, Lyle, and thanks for the invitation to come and present.
00:01:15.280 --> 00:01:16.799
I'm delighted to be here.
00:01:17.519 --> 00:01:37.599
Just before we kick off this afternoon, there's a quick poll that I wanted to kick off with for everyone that's on the on the call, just to get a gauge from you guys on how easy or difficult it is to recruit and to retain the right skills and talent across your teams.
00:01:38.000 --> 00:01:43.680
I'm assuming most people here will work in data and technology, and so keen to hear your views.
00:01:52.000 --> 00:01:53.120
Excellent.
00:01:53.439 --> 00:01:59.680
And yeah, and as I say, delighted to be here today to talk about all things data, tech, and talent, my favourite topics.
00:01:59.920 --> 00:02:06.560
And I thought I'd kick off today with a quick quote from Chat GPT.
00:02:06.959 --> 00:02:17.360
So when I was preparing for this webinar, um I asked ChatGPT how did it see the evolution of data skills across the last decade or so?
00:02:17.599 --> 00:02:20.639
And this is what it came back with, with no other prompt than that.
00:02:20.879 --> 00:02:32.159
So sales have expanded from specialized tech tools to a comprehensive ecosystem that includes engineering, deployment, ethics comms, and domain fluency.
00:02:32.639 --> 00:02:41.520
Organizations are investing in data democratization and literacy, while the labour market increasingly values adaptability over formal credentials.
00:02:41.680 --> 00:02:44.479
And I promise I didn't prompt it with adaptability.
00:02:44.639 --> 00:02:47.039
But that's what ChatGPT came back with.
00:02:47.360 --> 00:02:59.520
And I think for all of us who are working in the data and technology space, that first part of the sentence, we have seen over the last certainly couple of decades that I've worked in technology.
00:02:59.759 --> 00:03:08.879
We are doing the same things, but with an increasingly more sophisticated tools and the ability to do more with the change and growth in technology.
00:03:09.120 --> 00:03:20.400
But I think what we've seen in the last particularly five to ten years is just the acceleration of the tools and technology that we have available to us and how quickly that just moves.
00:03:20.639 --> 00:03:39.919
And in order to be able to take advantage of those new tools and tech and the opportunities that they offer, then it's essential to have a team who are flexible and who are adaptable to be able to relearn new skills and then to upskill in the tools and tech that are available to them.
00:03:40.080 --> 00:03:41.599
So yeah, I just wanted to start with that.
00:03:41.680 --> 00:03:47.039
Um so ChatGPT agrees to um we we value adaptability overall.
00:03:47.680 --> 00:03:49.439
Elizabeth, just sorry to interrupt.
00:03:49.599 --> 00:03:52.719
So 100% difficult came back from the poll.
00:03:53.120 --> 00:03:54.400
Which probably doesn't surprise you.
00:03:54.800 --> 00:03:55.759
It doesn't surprise me.
00:03:55.840 --> 00:04:03.919
I think it is uh Seth and Lauren touch on on some of that and my own personal experience with different domains um as I go through the talk today.
00:04:04.400 --> 00:04:04.960
Super.
00:04:05.199 --> 00:04:20.319
So before I start to talk about data skills and talent and what we need to do to kind of maintain the right kind of skills in the workforce, I wanted just to talk a wee bit just to touch on the evolution of data analysis itself.
00:04:20.399 --> 00:04:27.680
And that's really just to put into context and what that means for our skills that we need within the team.
00:04:28.480 --> 00:04:31.839
So all of these stories I'm certain will be familiar to you.
00:04:31.920 --> 00:04:40.240
But I wanted to start a way back in 1859 with Florence Nightingale and her radiant diagrams of cholera and how diseases spread.
00:04:40.480 --> 00:04:44.160
So super early example, there are many others, Napoleons, etc.
00:04:45.600 --> 00:04:51.199
O where we start to use data and visualization to understand patterns and to make decisions.
00:04:51.360 --> 00:04:54.160
So we've been doing this for a very, very long time.
00:04:55.439 --> 00:05:10.959
If we move on into the 1900s, into 1940, Alan Turing and breaking the enigma code during World War II, and really the first example of where we're starting to use big compute and to be able to solve those more complex statistical problems.
00:05:11.199 --> 00:05:23.360
So we've moved on from visualizing reams of data manually into how do we use big computers to be able to crack really complex problems, and we can program them to do that for us.
00:05:35.199 --> 00:05:44.720
Again, the first time where a computer was able to think critically and to use lots and lots and lots of rules to be able to beat a human being.
00:05:45.040 --> 00:05:49.759
And it really was a kind of pivotal moment in the tech world having the ability to do that.
00:05:50.560 --> 00:06:03.439
And then jumping to 2015, and Fan Wee, who was the world's leading alpha goal player, that super complex um puzzle game, and he was beaten by Deep Mind.
00:06:03.519 --> 00:06:14.480
So again, embedding deep learning, ML, AI, and the computer was able to beat him at one of the most complex or the most complex games in the world.
00:06:14.720 --> 00:06:23.759
There are lots and lots more examples that I'm sure that you're here all know over the last few years and certainly decades and centuries.
00:06:24.000 --> 00:06:32.319
But the reason that I wanted to put this up is to say that we have been doing data analysis for a very, very long time, for centuries, in fact.
00:06:32.560 --> 00:06:37.199
The way we have done data analysis has changed and evolved over time.
00:06:37.360 --> 00:06:42.959
And that's largely been driven by the tools and technology that we have available to us.
00:06:43.199 --> 00:07:06.399
And I think what we've really seen over the last couple of decades is the acceleration in the tools and tech that we can use, with the introduction, particularly of cloud computing and the ability to be able to run really large, complex models really quickly in real time, has just opened up a whole world of opportunity that we should take advantage of.
00:07:06.959 --> 00:07:12.800
So I wanted to start with that to say that actually we've been adapting, we've been changing over centuries and decades.
00:07:12.959 --> 00:07:21.839
I think this is just another period of change and one which our data and tech practitioners will continue to evolve into.
00:07:24.160 --> 00:07:27.360
So now I want to talk about a continually adapting team.
00:07:27.439 --> 00:07:33.519
So data analysis has evolved over time, and so have data teams in general.
00:07:34.160 --> 00:07:45.360
So what I wanted to touch on first of all are the types of roles and skills that are required across a data team and enabling them to do their job effectively.
00:07:45.519 --> 00:07:50.560
So I've just started with some of the really core roles that sit within the data team itself.
00:07:50.720 --> 00:08:05.519
So data engineering, bringing the data from whichever systems you own or have access to into a data lake, data warehouse, data mart, whatever it might be, to be able to carry out some data analyses.
00:08:06.240 --> 00:08:24.240
Historically, data engineering was carried out typically by BI developers and using ETL processes and has evolved in line with that evolution of technology and being able to build bespoke data engineering pipelines and think about the orchestration of data, particularly within cloud environments.
00:08:25.439 --> 00:08:26.319
Data analysis.
00:08:26.399 --> 00:08:31.920
So when we've got our data to where it needs to be, and what do we need to do to analyse that in either a simple or complex way?
00:08:32.000 --> 00:08:47.120
And that can cover all things from simple analysis, presenting that data through data visualization on the front end and for standard operational reporting, all the way into using the data for data science, machine learning, AI, etc.
00:08:48.480 --> 00:08:54.879
And I think the data analysis skill set itself can be one sometimes that's overlooked.
00:08:55.039 --> 00:09:09.600
That kind of business analyst, you know, simple analysis role, I think can add really key value into a team and being able to understand the business context and the data, and then being able to answer those kind of complex business questions.
00:09:10.960 --> 00:09:18.879
Then we've into data visualization, which is talked about, we've been doing this for decades and centuries, but now we can do it in a much more sophisticated way.
00:09:19.039 --> 00:09:32.960
So we've moved on in the last couple of decades from using PowerPoint and Excel to build out your bar charts and line charts into much more interactive tools through Power BI and other similar data biz platforms.
00:09:33.200 --> 00:09:48.000
I think the other evolution that we've seen though, particularly across data visualization, is the fact that a lot of these tools have moved from into low code or sometimes, in some cases, no code, like click, drag, drag and drop.
00:09:48.240 --> 00:10:06.960
So now our business analysts are able to pull together pretty good data visualizations with not very much coding language, and that's just really opened up the opportunity to be able to use those and also offered an opportunity for people to upskill and to start to utilize some of their skills through some of these database platforms.
00:10:07.840 --> 00:10:26.639
Then moving into data science, and again, that can be relatively simple data science, and from you know building out predictive models, using regression models, predicting what might happen in the future to some of our more complex models and touching on AI, using LLMs, deep learning, etc.
00:10:27.679 --> 00:10:49.519
And yeah, I think that's the area where we've really seen a lot of growth over the last decade or so and a lot of prevalence now in organizations where data science didn't always exist or existed in a very kind of small way in simple analysis, and we've now extended that out into building out MLAI models pretty much in most industries and most organisations.
00:10:50.559 --> 00:10:53.279
And then lastly, I've included a bucket for data governance.
00:10:53.440 --> 00:11:03.200
Nobody ever wants to talk about it, but it is really, really key to making sure that anything that you do build from your data can be utilized in decision making.
00:11:03.360 --> 00:11:13.519
The classic rubbish in, rubbish out, we need to make sure our data is governed, it's consistent, it's complete, etc., so that we are making decisions based on the right and complete set of information.
00:11:15.519 --> 00:11:21.759
I wanted then to touch on a couple of those which I see as actually being relatively new in the data space.
00:11:21.840 --> 00:11:25.440
Um, and first of all, touch on platform engineering.
00:11:25.679 --> 00:11:31.039
So platform engineering has existed for a very long time, particularly in the infrastructure space.
00:11:31.279 --> 00:11:38.159
But I think now we're seeing a higher demand for data-specific platform engineers.
00:11:38.320 --> 00:11:48.559
So people who know how to build infrastructure as a standard, who know how to scale models, who know how to optimize, how to maintain a low running cost for some of these things.
00:11:48.720 --> 00:11:58.240
And I think we're really starting to see that evolve as a skill set in its own right within the data team as opposed to something that could be thought about by the data engineers.
00:11:58.399 --> 00:12:02.240
And I kind of think that's where we were maybe five to ten years ago.
00:12:02.399 --> 00:12:18.000
Whereas now, given the level and the kind of scalability of cloud platforms for our data work, and platform engineering, I think, actually is a pretty important role, particularly for managing cost and scalability in the platforms.
00:12:18.960 --> 00:12:21.120
And then lastly, ML and AI engineering.
00:12:21.200 --> 00:12:31.840
So we've seen lots and lots of that um being talked about over the last few years, particularly, I think, with the release of Chat GPT to the general public way back, I think, in 2023.
00:12:32.000 --> 00:12:38.960
Um, it's become much more kind of common language now to talk about ML and AI, and where I don't think it was before.
00:12:39.200 --> 00:12:52.000
Um, and we're seeing then higher demand now for specific ML and AI engineering roles as opposed to just general data science, statistical kind of analysis roles.
00:12:52.080 --> 00:13:00.000
So it's really kind of evolving, and particularly thinking about MLOps as its own kind of skill set within its own right.
00:13:01.440 --> 00:13:03.440
So I'll pause there for just a second.
00:13:03.600 --> 00:13:11.519
Um, that's kind of my take on the types of roles and skills that you typically need across a modern data team.
00:13:11.759 --> 00:13:22.000
I think all of these roles, other than the couple of purple ones at the bottom, have existed over the last number of decades, but the tools and the tech that we are using have now changed and evolved.
00:13:22.320 --> 00:13:34.960
And I think um it'd be good now just to kind of understand from all of you guys where you see um, if at all, any skills gap across all of these roles um within the data team.
00:13:35.039 --> 00:13:38.639
So maybe just pause for a second and let you share that next poll.
00:14:20.720 --> 00:14:21.679
Technology.
00:14:22.399 --> 00:14:35.919
Right, so um ML AI engineering fifty-four percent, and then platform engineering data science, data virtualization, uh data analysis thirty-one percent, and data engineering fifteen percent.
00:14:36.080 --> 00:14:40.399
So probably no surprise in the ML space, but uh yeah, yeah.
00:14:40.480 --> 00:14:55.840
I think so and and yeah, it's uh it's actually really interesting because I think had I asked that same question, say five or so years ago, maybe five to ten years ago, I don't think platform engineering and ML engineering would even have really featured.
00:14:56.000 --> 00:14:59.279
And I think the biggest gap would have been within data engineering.
00:14:59.360 --> 00:15:11.279
And it's certainly something that I experienced um you know in recruiting for roles, and particularly in my my last role in industry in the GNECO, found it really difficult at a point in time to be able to recruit for data engineers.
00:15:11.360 --> 00:15:14.080
There was just too much demand and not enough supply.
00:15:14.240 --> 00:15:17.519
But what I think is just really interesting here is that that's obviously started to shift now.
00:15:17.600 --> 00:15:28.320
You know, we've started to get that kind of level of supply in the market, people have upskilled, reskilled, etc., within the data engineering space, and now we're seeing the next thing happening within AI engineering.
00:15:28.399 --> 00:15:35.039
So just think it's quite interesting, and it's almost an example of where the market moves with the technology.
00:15:35.200 --> 00:15:43.679
So I think that's just kind of called out really clearly that in actually in another five years' time, we are unlikely to see that same level of skills gap in ML and AI.
00:15:44.320 --> 00:15:44.639
Good.
00:15:44.720 --> 00:15:48.159
So thanks, thanks very much everyone for sharing your thoughts there.
00:15:48.639 --> 00:15:55.279
And as I mentioned, those are kind of the key roles that I would see across a successful data team.
00:15:55.519 --> 00:16:03.759
And actually, some of those roles might be the same person carrying out two roles and maybe smaller organizations with smaller data teams.
00:16:03.840 --> 00:16:15.600
You might have someone doing your engineering and your data science, may have someone else doing your data vis and your data analysis as a BA, but it's typically those kind of skill sets that we would look for across the team in general.
00:16:16.320 --> 00:16:37.279
But then to really make your data team successful, I think it's so important to remember the collaboration across the organization, both within your team itself and within the technology function, and secondly, across the actual organization who you're serving and supporting.
00:16:37.440 --> 00:16:45.039
Now, maybe just touch on an anecdote and taking that first path on collaborating across the technology space itself.
00:16:45.200 --> 00:16:58.000
And it's something actually I really experienced back in my time in Agreco and understanding the value of collaborating, of sharing ideas and of kind of building joint teams across the tech function.
00:16:58.960 --> 00:17:08.400
So while I was at Agreco, we implemented a modern data platform, so a single cloud-based platform to support all of our data work across the organization.
00:17:08.559 --> 00:17:14.880
That included sending messages from our AI and ML models to our in-house applications.
00:17:15.039 --> 00:17:24.640
It included producing Power BI or standard tabular extract data for our business, and it also enabled them to self-serve.
00:17:24.720 --> 00:17:32.000
So it really fed kind of all of the data needs that we had within the organization, centralized within this one platform.
00:17:32.720 --> 00:17:39.039
But in all honesty, we we struggled to implement it successfully over the first kind of 18 months.
00:17:39.200 --> 00:17:42.640
We'd aimed to have it in place, and around about 18 months, that was our project plan.
00:17:42.720 --> 00:17:49.839
But as we got to the kind of year point, we realized that actually it wasn't going to be in place, and we learned an awful lot along the way.
00:17:50.079 --> 00:18:11.200
But the real game changer for us and where we turned the program around and really accelerated the pace at which we were delivering is when we embedded one of our software developers within our data team, and we started to apply the principles of the software lifecycle development cycle.
00:18:11.359 --> 00:18:18.960
So, actually, we started to think about how do we think about this data platform almost as a software platform that we are then applying to data.
00:18:19.119 --> 00:18:26.799
And as soon as we started to apply those kind of rules and principles, and it was not overly complex at all, it was thinking about it in a different mindset.
00:18:26.960 --> 00:18:40.640
We really accelerated our delivery, we reduced down the number of defects, we got our code through much quicker, the quality was better, and it was just from thinking about blending all of our technology knowledge together in order to be able to get to that answer quickly.
00:18:40.799 --> 00:18:51.359
And I think that leads into what I was talking about earlier on with platform engineering and really thinking about the platform itself, its scalability, um the cost, being able to control that, the speed, etc.
00:18:51.680 --> 00:18:55.839
And really starting to collaborate across other parts of technology can help.
00:18:55.920 --> 00:19:02.880
And in really large organizations, you'll have a lot of opportunity to do that and by kind of tapping into some of the some of the apps teams, too.
00:19:02.960 --> 00:19:13.680
So that's just kind of one thing, and one thing that can really help also to to kind of upskill your data team and some of those tools, technologies, and processes too.
00:19:14.240 --> 00:19:19.119
And then the second thing I wanted to touch on in collaboration is just the collaboration with the business.
00:19:19.440 --> 00:19:28.319
So I'm not going to talk too much today about data strategy and how you make sure you're adding value and make that um make that impactful across the organization.
00:19:28.480 --> 00:19:39.279
Um but collaboration is absolutely key to make sure that first of all you understand the context of the data that you're working with, whether you're solving a simple or a complex problem.
00:19:39.599 --> 00:19:53.279
And then, second of all, that when you do develop a report, a dashboard, a model, that then the people within the organization utilize your output within their decision making because that is how we add value.
00:19:53.440 --> 00:20:02.400
So, although we're talking today mostly about data skills and how we make sure we're flexible, adaptable, and the collaboration across the organization is absolutely critical.
00:20:02.559 --> 00:20:15.680
And in every single one of these particular roles, communication and people skills are key because we need to make sure that whatever it is that we are creating is picked up, it's used, and it adds value to the organization.
00:20:17.440 --> 00:20:17.759
Okay.
00:20:18.640 --> 00:20:23.039
So I'm going to move on now a little bit to kind of recruiting and retaining talent.
00:20:23.200 --> 00:20:28.000
So when we asked at the poll at the start, you know, how do people find kind of trying to recruit for the right skills?
00:20:28.160 --> 00:20:29.279
What does that look like?
00:20:29.359 --> 00:20:31.039
Um, everyone said that was difficult.
00:20:31.119 --> 00:20:38.079
And I think we we do see that, you know, there's a demand for data and tech skills across the market, and our supply has not caught up generally.
00:20:38.880 --> 00:20:59.279
So, what I wanted to do now was just to talk through some examples of where I've had success and being able to recruit, and then really crucially retain talent as well to make sure that when you get those right people in the door, that actually they're a good fit for the organization, a good fit for the role, and you can then retain them for a reasonable period of time.
00:21:00.240 --> 00:21:03.119
So the first thing I wanted to touch on were apprentices.
00:21:03.599 --> 00:21:14.720
We're very lucky in the UK that we are offered an apprenticeship levy and in UK organisations, which means that the cost of taking apprentices on to organisations is very, very low.
00:21:15.200 --> 00:21:24.960
We had an apprenticeship programme specifically within the software development team within Agreco that we had absolutely huge success from.
00:21:25.200 --> 00:21:40.559
We brought in two or three apprentices per year, straight from school, and who joined our team and then they kind of cycled through a number of different roles within the software, the software dev team, and they added an absolutely huge amount of value.
00:21:40.799 --> 00:21:48.400
So I know sometimes people are apprehensive and want to recruit recruit experienced hires, they want people just to hit the ground running and to get go.
00:21:48.559 --> 00:21:59.759
But I promise you, investing in that kind of pyramid structure where you have a lot of resources at that junior level that you're training up, that you're upskilling, um, can add absolutely huge value to the organisation.
00:22:00.240 --> 00:22:11.279
And also brought some excellent ideas that actually some of our experienced team just hadn't thought about because they were coming onto things fresh phase, you know, brand new and asking kind of the right questions.
00:22:11.440 --> 00:22:16.559
So I would absolutely encourage you to have a think about apprenticeship programmes within your organization.
00:22:16.720 --> 00:22:26.319
And there are also lots of third parties that offer kind of data-specific training programmes through the apprenticeships who can then really support.
00:22:26.400 --> 00:22:35.119
So the apprentices will go off, you know, every couple of months, do some kind of external training with your third party and come back and apply some of that within the organization.
00:22:35.200 --> 00:22:47.440
So it's a really excellent scheme, and I would really encourage thinking about apprentices across data technology dev teams, and they can add a huge amount of value, and we had really positive experience with them.
00:22:48.640 --> 00:22:54.960
Next one I want to talk about is graduates, which is where I think lots of people do and tend to recruit for their teams.
00:22:55.039 --> 00:23:00.640
So, and again, we have done that across all of the roles and in the data teams that I've worked in.
00:23:00.880 --> 00:23:07.680
One thing I would say is that I've not always recruited from technical um degrees.
00:23:07.920 --> 00:23:14.079
So I've recruited people before into technical roles who studied arts degrees, for example.
00:23:14.319 --> 00:23:32.799
And I think that's a really I think that's something to kind of be very open-minded about, particularly when thinking about those data visualization um roles or the kind of data analysis type roles, and where it's kind of light touch on technical skills, but they tend to be very, very strong on communication skills.
00:23:32.880 --> 00:23:42.720
So it actually offers a really nice blended feel to the team and again a different way of thinking, which can only add strength to the overall team and how the team makes decisions.
00:23:42.799 --> 00:23:51.039
So yeah, graduates kind of standard, I'm sure most people use them, but I would encourage you to think about um recruiting from non-technical backgrounds as well as STEM.
00:23:52.559 --> 00:24:05.839
Postgraduate too, so again, thinking about people who have come out from PhDs or MSEs, etc., and bringing that a little bit more, usually deeper technical knowledge within the team.
00:24:06.160 --> 00:24:27.200
Um again, would recruit for those across all of the teams and typically into, as I said, more technical roles and with a particular skill set so that they've studied something like LLMs or something like that, if you want to be able to introduce that into your team, and they can bring their academic knowledge into that business context.
00:24:28.559 --> 00:24:43.519
And then the last one I kind of wanted to touch on, and related to the last poll that we asked, is where have people started to see or started to think about upskilling their team members and across the organization into some of these data roles.
00:24:43.839 --> 00:24:52.880
And the reason I wanted to touch on this is I mentioned previously that in Agreco we really struggled probably seven-ish years ago to recruit data engineers.
00:24:53.039 --> 00:24:55.440
We were implementing this cloud-based platform.
00:24:55.599 --> 00:25:05.759
We knew we needed dedicated data engineering support because our data scientists were doing the data engineering work, but we just could not recruit the people that we wanted.
00:25:06.079 --> 00:25:14.160
So after about six months of trying, um, we decided to instead just upskill the existing people that we had within our team.
00:25:14.319 --> 00:25:23.599
People who had worked as DBAs before, who'd worked as BI developers, who had complementary skill sets and whose roles were naturally evolving.
00:25:23.680 --> 00:25:28.079
We were moving away from you from using on-prem into using cloud.
00:25:28.160 --> 00:25:32.720
And so they wanted to transfer those skills to upskill and to give them further opportunity.
00:25:33.279 --> 00:25:44.720
And so, what we did instead is we essentially built a data engineering team with no experienced data engineers, but we had the opportunity to kind of upskill them as we implemented that cloud-based platform.
00:25:44.880 --> 00:25:49.200
And I'll talk a wee bit about how we did that in a minute when I talk about learning.
00:25:49.519 --> 00:26:03.920
Um, and I think the real benefit for us in doing that is that these people were already very technically competent, and they understood our in-house systems and the data and how it worked, and they understood the context of the data too.
00:26:04.079 --> 00:26:09.839
So, really, what we were doing is just laying on an additional technical skill for them to be able to use.
00:26:10.079 --> 00:26:21.759
And I think that's just a really good kind of tangible example of the world is changing, technology is changing, you know, the tools and tech we have access to are becoming more and more sophisticated.
00:26:21.839 --> 00:26:24.240
So, how do we just make sure that we keep up with that?
00:26:24.480 --> 00:26:35.359
Just iteratively adding on skills and competencies as we go and to make sure we're kind of getting best benefit and that people are kind of evolving as the as the tools and the tech do as well.
00:26:37.200 --> 00:26:49.920
And then the last thing I wanted to touch on in recruiting and retaining talent is just being able to tap in and access government or vendor-funded programs to support and enhance innovation.
00:26:50.319 --> 00:26:55.039
And I just touched on a couple of examples of again where we did that within a GRECO.
00:26:55.599 --> 00:27:05.839
So many of you in Scotland will be familiar with the Data Lab and who support a number of students to be able to study data science and related topics across Scotland.
00:27:06.240 --> 00:27:18.000
And we supported a number of the data science graduates in their MSc programmes, and then went on to recruit some of those graduates after the MSc programmes within the team.
00:27:18.240 --> 00:27:24.319
And that was a really good way for us to be able to, first of all, test out some new and innovative ideas.
00:27:24.400 --> 00:27:41.359
We had over a six to eight week period of them carrying out their project with us, they got a bit of a flavour of what the organization was like, and so it was a really easy transition then when they completed their MSc and to offer them a role within the team, and they came and we had a great success of a number of people coming to join us.
00:27:41.440 --> 00:27:56.000
So it's a really good way of being able to kind of tap into different talent and also through those kind of MSc projects, six to eight weeks to test out some of the innovative ideas you have in the team, but without putting at risk any of the other operational delivery.
00:27:57.279 --> 00:28:01.599
And then the last thing I wanted to touch on was the was working with Innovate UK.
00:28:01.759 --> 00:28:06.240
So there's a number of government bodies that offer funding for innovation.
00:28:06.319 --> 00:28:08.559
So Innovate UK is one of them.
00:28:08.799 --> 00:28:22.799
You can apply for funding over a period of time, and if that is granted, you then work collaboratively with academia and with someone who is sponsored by Innovate UK to tackle an innovative problem that you have within the organization.
00:28:23.039 --> 00:28:32.480
And that again is just a really fantastic way of being able to, at very low risk and low cost, test out some innovative ideas that you have within the team.
00:28:32.640 --> 00:28:38.799
Maybe you've never tried to implement a GenTic AI before and you want to try that, you want to know what it might look like.
00:28:38.960 --> 00:28:54.160
So you come up with a kind of two or three year programme where you're going to do that, you select a university you're going to work with, it gives you that collaboration between public sector, private sector, and academia, and essentially delivers great value at very low risk for the organization.
00:28:54.319 --> 00:29:07.359
So if you're not doing so already, I would absolutely encourage you to have a think about any of those schemes and think about how you can utilize them, as I say, mostly for low risk and low cost to be able to test out some of new innovative ideas.
00:29:09.279 --> 00:29:26.799
Conscious of time, I'm going to say a couple more slides now in terms of keeping skills relevant and when you've got that team in place, what do you do to be able to maintain um to keep them within the team and for them to be able to maintain their skill set and to grow that through their time with you?
00:29:27.599 --> 00:29:50.160
And it's something we used to talk about quite a lot, actually, across all of the kind of data teams that I've worked in, where I think particularly at more junior grades, they were very keen to make sure that there was dedicated time for learning, what the learning program would be, how they were going to grow in the role, how they were going to develop, and wanted kind of very concrete answers to some of that.
00:29:50.240 --> 00:29:53.920
So I'm going to talk through how I kind of typically approach that.
00:29:54.160 --> 00:30:01.680
And often a lot of the ask was can we have structured time and dedicated time every week to be able to dedicate to learning?
00:30:01.839 --> 00:30:07.119
And so I'm quite keen to understand from this group on the call, um, is that something that you typically offer?
00:30:07.359 --> 00:30:10.240
Is that difficult to find time for within the organization?
00:30:10.319 --> 00:30:13.119
Um, so we can pop that next pull up, that would be great.
00:30:39.039 --> 00:30:42.880
So eighty eighty-seven percent know we struggle with that question.
00:30:43.599 --> 00:30:46.079
And and I think that's that's quite typical, isn't it?
00:30:46.160 --> 00:30:50.960
Like in organizations, there's always demand, there's always a deadline, there's always something that's really urgent that comes up.
00:30:51.119 --> 00:31:00.720
And it's always really easy to drop that thing which is kind of additive, it's learning, and then it might not be directly relevant to the deadline or the thing you're trying to deliver next week.
00:31:00.960 --> 00:31:11.920
And it definitely has been a challenge in teams that I've worked across and people just really wanting that kind of dedicated time and it being difficult to try and always, you know, ring fence those couple of hours a week or whatever it might be.
00:31:12.160 --> 00:31:27.039
Um so what I started to do was to encourage teams to think about learning in four different ways and to help them to reflect on how and when they are learning, particularly when that specific or bespoke time that was ring fenced just doesn't happen.
00:31:27.440 --> 00:31:31.359
So the first one I would talk about is on the job learning.
00:31:31.519 --> 00:31:35.119
So I've talked quite a lot about how data and tech world has evolved.
00:31:35.279 --> 00:31:37.920
There is no option but to learn on the job.
00:31:38.079 --> 00:31:47.440
You are either learning a new skill, you might be learning, you might already use Python, but actually you're going to use a new Python library for the new pro problem you're trying to solve.
00:31:47.759 --> 00:31:54.400
You might be working on a new business problem with new data and a new context that you haven't previously understood.
00:31:54.559 --> 00:32:05.599
And I think helping your teams to recognize that that is also learning, helping them to reflect through their personal development plans on what they know now versus what they knew six months ago.
00:32:05.920 --> 00:32:13.359
I think just helping people to understand that when we work in an industry like data and technology, you have to learn just a standard.
00:32:13.440 --> 00:32:18.160
So on-the-job learning absolutely counts and absolutely happens across all data teams.
00:32:18.240 --> 00:32:19.839
I have no doubt of that.
00:32:21.359 --> 00:32:26.640
The second one I think is the one that we're often asked about, which is like independent, or I've said here, or group study.
00:32:26.880 --> 00:32:31.519
So where possible, having weekly focus time to be able to meet some goals.
00:32:31.680 --> 00:32:42.400
And I think in particular, it's it's always helpful to target that against wanting to get, for example, another Microsoft certification or wanting to complete a course on Udemy or Coursera.
00:32:42.720 --> 00:32:50.960
So wherever I've granted that kind of weekly focus time to say, you know, there's two hours on a Friday afternoon, you know, spend it like this, it's with a target in mind.
00:32:51.119 --> 00:32:58.400
It's within six weeks' time, we want to have completed this course, we want to have completed this cert, and we're looking for an outcome at the end of it.
00:32:58.480 --> 00:33:00.880
So it's a really targeted and focused time.
00:33:01.519 --> 00:33:06.559
The other thing is that outcome should obviously in some way be related to the role that they're doing.
00:33:06.640 --> 00:33:15.920
So if someone in the team came to me and said, We used to use Azure in my previous role, I want to go and do this course in AWS, the answer would likely be that's not particularly relevant.
00:33:16.000 --> 00:33:22.559
Um so it should be focused on something that they're working on or something that will be of benefit in the near future.
00:33:23.440 --> 00:33:29.200
The other couple of things we that I've kind of explored in the past which have worked quite well are regular tutorials.
00:33:29.359 --> 00:33:35.759
So typically bringing people in and across your data and tech teams come from different backgrounds, might have studied different subjects.
00:33:35.920 --> 00:33:48.960
So getting people to carry out a tutorial once every two weeks, once every four weeks on something that they are an expert in, and that might actually be presenting back the output of the model they're currently working on within the business context.
00:33:49.119 --> 00:34:00.880
But just setting it up like a tutorial and really framing it like that because it's a learning experience really helps people to kind of link that learning with their role and kind of feel like they're they're still developing.
00:34:01.599 --> 00:34:03.599
Then the last thing is kind of hackathons.
00:34:03.680 --> 00:34:07.200
So I've included that within like groups, group study, if you like, as well.
00:34:07.359 --> 00:34:20.159
Um, and actually I've run a number of hackathons with peer groups, so there might be other organizations who are similar to yours with similar teams, tools, and tech, and getting together, setting up a problem, and then letting people go on and solve that.
00:34:20.239 --> 00:34:33.440
And so taking one day every six months or so and getting together to do that, and again, that just kind of really helps, I think, with sometimes with team building and with upskilling and really helping people to promote additional learning.
00:34:34.880 --> 00:34:37.440
Third thing in terms of learning is classroom-based.
00:34:37.599 --> 00:34:47.840
So I've typically employed that when we have implemented a new tool or technology that we need to learn about, and we want some dedicated and bespoke time on that from our vendors.
00:34:48.320 --> 00:34:59.599
So those could be through people like you know AWS, Google, Microsoft, Databricks, etc., dedicated sessions on their tool and tech and how to get the most out of it.
00:34:59.920 --> 00:35:03.599
Or the classroom-based learning could be on people skills.
00:35:03.679 --> 00:35:11.039
So we've got the output of our model, how do we influence the business, and really thinking about how to bring people in to support in that domain too.
00:35:11.199 --> 00:35:15.599
So another type of kind of learning that people can have access to.
00:35:16.719 --> 00:35:20.320
And then the last thing, um, conferences and peer connections.
00:35:20.400 --> 00:35:36.960
Um and I would always encourage people at every single stage of their career to keep up with their peers that they met at any stage and to keep in touch because you will always learn, be able to bounce ideas off each other and learn something new from what it is that they're doing.
00:35:37.199 --> 00:35:39.519
There are lots of local meetups across Scotland.
00:35:39.679 --> 00:35:42.480
Data Scotland is coming up, I think, again in September.
00:35:42.559 --> 00:35:50.239
Um, international events like Big Data London, um, Data AI Summit, etc., and lots of vendor-led forums.
00:35:50.320 --> 00:35:58.880
Um either recruiters tend to run lots of kind of roundtables, um dinners, meetups, etc., as do software providers.
00:35:59.039 --> 00:36:09.280
So I would just encourage people to think about what they have access to, either remotely, webinars like this one, and or in person, and where there are meetups that happen regularly.
00:36:09.679 --> 00:36:16.320
So yeah, I would just kind of I think that's always a challenge trying to carve out time for learning to keep those skills up to date.
00:36:16.480 --> 00:36:26.079
But those are the four ways that I would really think about learning reskilling and upskilling that can absolutely be embedded with relative ease within your day-to-day job.
00:36:27.119 --> 00:36:28.880
Conscious of time, last slide.
00:36:29.039 --> 00:36:40.800
Um, to finish on, I just wanted to pop on some helpful sources for leaders, and just in my experience, where I've really found it beneficial and to kind of keep up with the market and what's what's going on.
00:36:41.039 --> 00:36:43.280
The first, again, as I said, networking with peers.
00:36:43.360 --> 00:36:58.239
There are lots of senior data groups and roundtables, um, either that work remotely or in person, and those are usually invaluable for being able to share ideas, share what went well, what didn't go well, and learnings.
00:36:58.320 --> 00:37:05.679
Um so, yeah, that's probably my number one um tip is to always keep up with peers and what they are doing.
00:37:05.760 --> 00:37:09.440
Um it really is invaluable to learn from people who are doing the same thing as you.
00:37:10.159 --> 00:37:11.679
Maximising vendor support.
00:37:11.760 --> 00:37:19.199
Um, vendors often have some budget to be able to help you to do some of the things that you want to do, um, either in small projects or otherwise.
00:37:19.360 --> 00:37:21.440
Um so please do lean on vendors.
00:37:21.519 --> 00:37:27.840
I think there are lots of organizations who don't probably do that enough, and they're always quite keen to offer help and support.
00:37:27.920 --> 00:37:39.119
They'll often try to sell you something too, and but usually you can get good learning from them, and also it helps you to keep up with what's coming up next in the market and what other tools or tech might be helpful for your team.
00:37:39.760 --> 00:37:58.079
I plopped in your consultancy support, and I think that one for me is is quite an important one, particularly when trying something new, because consultancies have the benefit that they have already largely done the thing you're trying to do with other organizations and they've seen where it went well or where it didn't.
00:37:58.239 --> 00:38:16.320
So whether you engage with you know a third party to implement something, to get some advisory support, or even just to kind of keep up with what's going on, I would just always kind of keep some kind of connection to consultancies, particularly that operate in the data space, and to kind of keep ahead of where the market's at and what opportunities there might be.
00:38:16.719 --> 00:38:25.679
And then lastly, linked to that, and something I found hugely helpful in my time in Agreco was utilising the Gartner GTP license for technical professionals.
00:38:25.840 --> 00:38:30.480
And for me, that was really helpful, particularly when we were implementing the modern data platform.
00:38:30.639 --> 00:38:36.239
We knew what we wanted to do, but we were pretty agnostic about some of the tools and tech that we wanted to use.
00:38:36.480 --> 00:38:39.679
You can't ask the vendors those questions because their tool is always the best.
00:38:39.760 --> 00:38:47.199
Um so really kind of leaned on Gartner when we had short lists of vendors and saying, here's the things we're considering for these different um functionalities.
00:38:47.360 --> 00:38:48.000
What do you think?
00:38:48.159 --> 00:38:49.199
What would your advice be?
00:38:49.360 --> 00:38:50.880
What's going to work best for us or not?
00:38:50.960 --> 00:38:52.800
And we just got some really good sound advice.
00:38:52.880 --> 00:39:00.079
So for me, that was just really helpful, particularly when making big technology decisions to get a kind of software diagnostic view.
00:39:01.360 --> 00:39:04.480
So um that's been that's all I wanted to talk through today.
00:39:04.559 --> 00:39:16.000
I hope that that's been helpful and just have a talk through current data teams, skills, tools, technologies, and what you can do to maintain that high-performing um data team across the organization.
00:39:17.039 --> 00:39:17.360
Great.
00:39:17.519 --> 00:39:19.039
Thank you so much, Elizabeth.
00:39:19.360 --> 00:39:20.159
Thank you for that.
00:39:20.480 --> 00:39:23.519
Lots of uh hints and tips we can all take back to our visitors.
00:39:23.679 --> 00:39:24.559
I really appreciate it.
00:39:24.719 --> 00:39:27.760
I am cautious of time, so I'm just gonna go straight into questions.
00:39:27.920 --> 00:39:33.280
Um there is a couple of questions with regards to AI and the replacement of roles.
00:39:33.440 --> 00:39:40.559
So I'm talking specifically within kind of data teams and contact with adoption of AI within lower level roles.
00:39:40.719 --> 00:39:45.440
I know you spoke about uh kind of entry-level opportunities and new roles for graduates.
00:39:45.519 --> 00:39:47.519
How do you think AI is going to impact that?
00:39:48.480 --> 00:39:50.320
Yeah, no, I think it's actually interesting.
00:39:50.559 --> 00:39:58.400
I think for me generally, my view on AI is always this it's going to enhance and augment human decision making.
00:39:58.639 --> 00:40:02.000
So I don't think we're going to see AI take over everything.
00:40:02.159 --> 00:40:08.559
I do think we'll see the replacement of some of those kind of, you know, roles which can be automated.
00:40:08.639 --> 00:40:17.679
Um and I think we're seeing that already sometimes, you know, we're no longer right, and we've seen this for years, we're no longer coding out particular statistical models from scratch.
00:40:17.760 --> 00:40:19.199
You know, we pick up a Python library.
00:40:19.280 --> 00:40:20.800
We've been doing that for a very long time.
00:40:20.960 --> 00:40:24.079
So I think we'll start to see those kind of iterative things happening.
00:40:24.320 --> 00:40:33.920
Um I think yes, it has the there is, you know, the risk that it affects those kind of more junior roles because you automate some of those things.
00:40:34.159 --> 00:40:42.639
But I also think that organizations should be savvy in understanding that critical thinking is still a really important skill to have.
00:40:42.800 --> 00:40:51.920
And I think if we automate too much and too early, then people don't really understand all of the things that they are automating and can't really make kind of judgments based on that.
00:40:52.000 --> 00:40:57.360
Um, so I think we'll still see the opportunity for people to be able to do some of those roles.
00:40:57.440 --> 00:41:03.039
I don't think everything will be automated, and I still think we'll just still see a demand for any of the data and tech roles.
00:41:03.199 --> 00:41:05.760
Like I just don't see it reducing kind of over time.
00:41:06.320 --> 00:41:06.639
Okay.
00:41:06.800 --> 00:41:08.320
Okay, no, thanks for that.
00:41:08.480 --> 00:41:14.159
Um, I think following on from that, you've kind of answered that question, but a little bit more on specifically the future.
00:41:14.239 --> 00:41:22.320
So within data and AI, where do you see the kind of most in-demand skill sets then within you know those areas moving forward?
00:41:22.880 --> 00:41:25.360
Yeah, so I think we we touched on it earlier on.
00:41:25.519 --> 00:41:31.519
I think the the demand and the shortage we're seeing just now is that kind of platform engineering and ML AI engineering space.
00:41:31.679 --> 00:41:35.039
I just don't think that we're quite caught up yet with the supply on that.
00:41:35.280 --> 00:41:43.199
I think there's a huge opportunity for people to upskill who are already working in data to be able to operate and start to move within that space.
00:41:43.519 --> 00:41:57.760
But the other thing I think that we absolutely can see, and I think, and I touched on it earlier too, is that people need to have the right kind of people and soft skills to be able to actually get the benefit of these AI models.
00:41:57.920 --> 00:42:03.760
People are always a bit frightened of what might be a black box, like how should we implement it, what is that going to mean.
00:42:03.920 --> 00:42:17.199
So I think really being able to explain what a model is doing under the hood at a very high level, how it can be implemented, what that value might be is really, really key to making sure you have a really kind of successful data team.
00:42:17.280 --> 00:42:24.960
So I think the two shortages for me on the tech side, look at that kind of ML, AI engineering, MLOps, how do you make sure your model's not going to drift?
00:42:25.039 --> 00:42:26.719
How do you make sure you can scale it in the right way?
00:42:26.880 --> 00:42:30.079
How do you make sure costs just don't become you know out of control?
00:42:30.239 --> 00:42:40.239
Um and then on the other side, actually, that kind of influencing, continuing to kind of promote the use of AI and help people to understand it and not be frightened of it too.
00:42:40.320 --> 00:42:41.440
I think there's those two elements.
00:42:41.920 --> 00:42:42.159
Okay.
00:42:42.239 --> 00:42:44.960
No, that's a good answer and uh lots to take away there.
00:42:45.039 --> 00:42:45.679
So thanks for that.
00:42:45.840 --> 00:42:47.599
I think we've got time for one more question.
00:42:47.760 --> 00:43:00.320
So somebody's um asked, how would you advise someone going about challenging or influencing instances where data is produced but potentially being ignored, misused, misinterpreted by management specifically?
00:43:00.800 --> 00:43:01.440
Yeah.
00:43:02.239 --> 00:43:06.559
Um I think it's a good question, and I think it's something that's just very common.
00:43:06.719 --> 00:43:11.119
Um, people often torture the data to get to the answer that they want to see.
00:43:11.280 --> 00:43:13.920
And for me, that's really just a question of leadership.
00:43:14.079 --> 00:43:19.119
Um, so there needs to be the right leadership in place to be able to promote the right use.
00:43:19.199 --> 00:43:32.079
And again, that comes back to the influencing, the tools, the tech, being able to explain what it is that they are doing so that people understand what's happening, and then they take the answer as being the accepted answer and not just something they want to torture to get to the right thing.
00:43:32.159 --> 00:43:38.800
So for me, that's just all about leadership and making sure that the person who understands the model has a seat at the table and is listened to.
00:43:39.280 --> 00:43:39.599
Okay.
00:43:40.000 --> 00:43:40.400
Great.
00:43:40.559 --> 00:43:41.840
Thank you, Elizabeth.
00:43:42.000 --> 00:43:44.159
Um I'm afraid to say that time is up.
00:43:44.239 --> 00:43:47.599
So I just wanted to say thank you to everyone for joining the call today.
00:43:47.760 --> 00:43:50.880
Massive thanks to Elizabeth for delivering that presentation.
00:43:51.039 --> 00:43:55.119
Lots to take away um that we can all use into our businesses moving forward.
00:43:55.199 --> 00:43:56.480
So I really do appreciate it.
00:43:56.719 --> 00:43:57.039
Thank you.
00:43:57.519 --> 00:44:08.800
Um just finally from me, I We'll be hosting our final Dayton AI webinar um in a couple of weeks with Callum Sinclair from Burness Paul talking everything AI and legal.
00:44:08.880 --> 00:44:11.360
So look forward to seeing you on the call then.
00:44:11.519 --> 00:44:11.760
Okay.
00:44:11.920 --> 00:44:12.880
Have a great day, guys.
00:44:12.960 --> 00:44:13.280
Thank you.
00:44:13.360 --> 00:44:13.920
Bye bye.
00:44:14.079 --> 00:44:14.639
Thanks, Al.
00:00:00.000 --> 00:00:06.799
Hello everybody, welcome to the fourth episode of our Talking Data and AI webinar series.
00:00:06.959 --> 00:00:10.800
My name is Lyle Richie, and I'll be your host today.
00:00:11.279 --> 00:00:15.199
Today we'll explore evolving skills and the data driven world.
00:00:16.079 --> 00:00:19.280
I'm delighted to introduce today's guest, Elizabeth Hollinger.
00:00:19.519 --> 00:00:29.760
Elizabeth has over 20 years' experience in data and how data has budget roles across a range of companies, including Deloitte, Avanti, and Graphics.
00:00:30.079 --> 00:00:34.880
So you could say to note a thing or two about reading high performance data and tech teams.
00:00:35.439 --> 00:00:41.280
And I'll just talk about that in more detail and about how to build adaptability as a course culture.
00:00:42.560 --> 00:00:45.920
So before I hand over to Elizabeth, just a couple of quick points.
00:00:46.640 --> 00:00:54.079
Elizabeth will talk to all today for around 30 minutes, and then there will be 50 minutes for a QA.
00:00:54.479 --> 00:01:04.799
So if you could please use the QA box and the chat function, and if you could thumbs up the questions you would like the answers to, and we will get those answered for you.
00:01:05.120 --> 00:01:09.439
So really looking forward to this, and I will hand over to Elizabeth.
00:01:09.680 --> 00:01:10.879
Elizabeth, over to you.
00:01:11.599 --> 00:01:15.200
Thanks very much, Lyle, and thanks for the invitation to come and present.
00:01:15.280 --> 00:01:16.799
I'm delighted to be here.
00:01:17.519 --> 00:01:37.599
Just before we kick off this afternoon, there's a quick poll that I wanted to kick off with for everyone that's on the on the call, just to get a gauge from you guys on how easy or difficult it is to recruit and to retain the right skills and talent across your teams.
00:01:38.000 --> 00:01:43.680
I'm assuming most people here will work in data and technology, and so keen to hear your views.
00:01:52.000 --> 00:01:53.120
Excellent.
00:01:53.439 --> 00:01:59.680
And yeah, and as I say, delighted to be here today to talk about all things data, tech, and talent, my favourite topics.
00:01:59.920 --> 00:02:06.560
And I thought I'd kick off today with a quick quote from Chat GPT.
00:02:06.959 --> 00:02:17.360
So when I was preparing for this webinar, um I asked ChatGPT how did it see the evolution of data skills across the last decade or so?
00:02:17.599 --> 00:02:20.639
And this is what it came back with, with no other prompt than that.
00:02:20.879 --> 00:02:32.159
So sales have expanded from specialized tech tools to a comprehensive ecosystem that includes engineering, deployment, ethics comms, and domain fluency.
00:02:32.639 --> 00:02:41.520
Organizations are investing in data democratization and literacy, while the labour market increasingly values adaptability over formal credentials.
00:02:41.680 --> 00:02:44.479
And I promise I didn't prompt it with adaptability.
00:02:44.639 --> 00:02:47.039
But that's what ChatGPT came back with.
00:02:47.360 --> 00:02:59.520
And I think for all of us who are working in the data and technology space, that first part of the sentence, we have seen over the last certainly couple of decades that I've worked in technology.
00:02:59.759 --> 00:03:08.879
We are doing the same things, but with an increasingly more sophisticated tools and the ability to do more with the change and growth in technology.
00:03:09.120 --> 00:03:20.400
But I think what we've seen in the last particularly five to ten years is just the acceleration of the tools and technology that we have available to us and how quickly that just moves.
00:03:20.639 --> 00:03:39.919
And in order to be able to take advantage of those new tools and tech and the opportunities that they offer, then it's essential to have a team who are flexible and who are adaptable to be able to relearn new skills and then to upskill in the tools and tech that are available to them.
00:03:40.080 --> 00:03:41.599
So yeah, I just wanted to start with that.
00:03:41.680 --> 00:03:47.039
Um so ChatGPT agrees to um we we value adaptability overall.
00:03:47.680 --> 00:03:49.439
Elizabeth, just sorry to interrupt.
00:03:49.599 --> 00:03:52.719
So 100% difficult came back from the poll.
00:03:53.120 --> 00:03:54.400
Which probably doesn't surprise you.
00:03:54.800 --> 00:03:55.759
It doesn't surprise me.
00:03:55.840 --> 00:04:03.919
I think it is uh Seth and Lauren touch on on some of that and my own personal experience with different domains um as I go through the talk today.
00:04:04.400 --> 00:04:04.960
Super.
00:04:05.199 --> 00:04:20.319
So before I start to talk about data skills and talent and what we need to do to kind of maintain the right kind of skills in the workforce, I wanted just to talk a wee bit just to touch on the evolution of data analysis itself.
00:04:20.399 --> 00:04:27.680
And that's really just to put into context and what that means for our skills that we need within the team.
00:04:28.480 --> 00:04:31.839
So all of these stories I'm certain will be familiar to you.
00:04:31.920 --> 00:04:40.240
But I wanted to start a way back in 1859 with Florence Nightingale and her radiant diagrams of cholera and how diseases spread.
00:04:40.480 --> 00:04:44.160
So super early example, there are many others, Napoleons, etc.
00:04:45.600 --> 00:04:51.199
O where we start to use data and visualization to understand patterns and to make decisions.
00:04:51.360 --> 00:04:54.160
So we've been doing this for a very, very long time.
00:04:55.439 --> 00:05:10.959
If we move on into the 1900s, into 1940, Alan Turing and breaking the enigma code during World War II, and really the first example of where we're starting to use big compute and to be able to solve those more complex statistical problems.
00:05:11.199 --> 00:05:23.360
So we've moved on from visualizing reams of data manually into how do we use big computers to be able to crack really complex problems, and we can program them to do that for us.
00:05:35.199 --> 00:05:44.720
Again, the first time where a computer was able to think critically and to use lots and lots and lots of rules to be able to beat a human being.
00:05:45.040 --> 00:05:49.759
And it really was a kind of pivotal moment in the tech world having the ability to do that.
00:05:50.560 --> 00:06:03.439
And then jumping to 2015, and Fan Wee, who was the world's leading alpha goal player, that super complex um puzzle game, and he was beaten by Deep Mind.
00:06:03.519 --> 00:06:14.480
So again, embedding deep learning, ML, AI, and the computer was able to beat him at one of the most complex or the most complex games in the world.
00:06:14.720 --> 00:06:23.759
There are lots and lots more examples that I'm sure that you're here all know over the last few years and certainly decades and centuries.
00:06:24.000 --> 00:06:32.319
But the reason that I wanted to put this up is to say that we have been doing data analysis for a very, very long time, for centuries, in fact.
00:06:32.560 --> 00:06:37.199
The way we have done data analysis has changed and evolved over time.
00:06:37.360 --> 00:06:42.959
And that's largely been driven by the tools and technology that we have available to us.
00:06:43.199 --> 00:07:06.399
And I think what we've really seen over the last couple of decades is the acceleration in the tools and tech that we can use, with the introduction, particularly of cloud computing and the ability to be able to run really large, complex models really quickly in real time, has just opened up a whole world of opportunity that we should take advantage of.
00:07:06.959 --> 00:07:12.800
So I wanted to start with that to say that actually we've been adapting, we've been changing over centuries and decades.
00:07:12.959 --> 00:07:21.839
I think this is just another period of change and one which our data and tech practitioners will continue to evolve into.
00:07:24.160 --> 00:07:27.360
So now I want to talk about a continually adapting team.
00:07:27.439 --> 00:07:33.519
So data analysis has evolved over time, and so have data teams in general.
00:07:34.160 --> 00:07:45.360
So what I wanted to touch on first of all are the types of roles and skills that are required across a data team and enabling them to do their job effectively.
00:07:45.519 --> 00:07:50.560
So I've just started with some of the really core roles that sit within the data team itself.
00:07:50.720 --> 00:08:05.519
So data engineering, bringing the data from whichever systems you own or have access to into a data lake, data warehouse, data mart, whatever it might be, to be able to carry out some data analyses.
00:08:06.240 --> 00:08:24.240
Historically, data engineering was carried out typically by BI developers and using ETL processes and has evolved in line with that evolution of technology and being able to build bespoke data engineering pipelines and think about the orchestration of data, particularly within cloud environments.
00:08:25.439 --> 00:08:26.319
Data analysis.
00:08:26.399 --> 00:08:31.920
So when we've got our data to where it needs to be, and what do we need to do to analyse that in either a simple or complex way?
00:08:32.000 --> 00:08:47.120
And that can cover all things from simple analysis, presenting that data through data visualization on the front end and for standard operational reporting, all the way into using the data for data science, machine learning, AI, etc.
00:08:48.480 --> 00:08:54.879
And I think the data analysis skill set itself can be one sometimes that's overlooked.
00:08:55.039 --> 00:09:09.600
That kind of business analyst, you know, simple analysis role, I think can add really key value into a team and being able to understand the business context and the data, and then being able to answer those kind of complex business questions.
00:09:10.960 --> 00:09:18.879
Then we've into data visualization, which is talked about, we've been doing this for decades and centuries, but now we can do it in a much more sophisticated way.
00:09:19.039 --> 00:09:32.960
So we've moved on in the last couple of decades from using PowerPoint and Excel to build out your bar charts and line charts into much more interactive tools through Power BI and other similar data biz platforms.
00:09:33.200 --> 00:09:48.000
I think the other evolution that we've seen though, particularly across data visualization, is the fact that a lot of these tools have moved from into low code or sometimes, in some cases, no code, like click, drag, drag and drop.
00:09:48.240 --> 00:10:06.960
So now our business analysts are able to pull together pretty good data visualizations with not very much coding language, and that's just really opened up the opportunity to be able to use those and also offered an opportunity for people to upskill and to start to utilize some of their skills through some of these database platforms.
00:10:07.840 --> 00:10:26.639
Then moving into data science, and again, that can be relatively simple data science, and from you know building out predictive models, using regression models, predicting what might happen in the future to some of our more complex models and touching on AI, using LLMs, deep learning, etc.
00:10:27.679 --> 00:10:49.519
And yeah, I think that's the area where we've really seen a lot of growth over the last decade or so and a lot of prevalence now in organizations where data science didn't always exist or existed in a very kind of small way in simple analysis, and we've now extended that out into building out MLAI models pretty much in most industries and most organisations.
00:10:50.559 --> 00:10:53.279
And then lastly, I've included a bucket for data governance.
00:10:53.440 --> 00:11:03.200
Nobody ever wants to talk about it, but it is really, really key to making sure that anything that you do build from your data can be utilized in decision making.
00:11:03.360 --> 00:11:13.519
The classic rubbish in, rubbish out, we need to make sure our data is governed, it's consistent, it's complete, etc., so that we are making decisions based on the right and complete set of information.
00:11:15.519 --> 00:11:21.759
I wanted then to touch on a couple of those which I see as actually being relatively new in the data space.
00:11:21.840 --> 00:11:25.440
Um, and first of all, touch on platform engineering.
00:11:25.679 --> 00:11:31.039
So platform engineering has existed for a very long time, particularly in the infrastructure space.
00:11:31.279 --> 00:11:38.159
But I think now we're seeing a higher demand for data-specific platform engineers.
00:11:38.320 --> 00:11:48.559
So people who know how to build infrastructure as a standard, who know how to scale models, who know how to optimize, how to maintain a low running cost for some of these things.
00:11:48.720 --> 00:11:58.240
And I think we're really starting to see that evolve as a skill set in its own right within the data team as opposed to something that could be thought about by the data engineers.
00:11:58.399 --> 00:12:02.240
And I kind of think that's where we were maybe five to ten years ago.
00:12:02.399 --> 00:12:18.000
Whereas now, given the level and the kind of scalability of cloud platforms for our data work, and platform engineering, I think, actually is a pretty important role, particularly for managing cost and scalability in the platforms.
00:12:18.960 --> 00:12:21.120
And then lastly, ML and AI engineering.
00:12:21.200 --> 00:12:31.840
So we've seen lots and lots of that um being talked about over the last few years, particularly, I think, with the release of Chat GPT to the general public way back, I think, in 2023.
00:12:32.000 --> 00:12:38.960
Um, it's become much more kind of common language now to talk about ML and AI, and where I don't think it was before.
00:12:39.200 --> 00:12:52.000
Um, and we're seeing then higher demand now for specific ML and AI engineering roles as opposed to just general data science, statistical kind of analysis roles.
00:12:52.080 --> 00:13:00.000
So it's really kind of evolving, and particularly thinking about MLOps as its own kind of skill set within its own right.
00:13:01.440 --> 00:13:03.440
So I'll pause there for just a second.
00:13:03.600 --> 00:13:11.519
Um, that's kind of my take on the types of roles and skills that you typically need across a modern data team.
00:13:11.759 --> 00:13:22.000
I think all of these roles, other than the couple of purple ones at the bottom, have existed over the last number of decades, but the tools and the tech that we are using have now changed and evolved.
00:13:22.320 --> 00:13:34.960
And I think um it'd be good now just to kind of understand from all of you guys where you see um, if at all, any skills gap across all of these roles um within the data team.
00:13:35.039 --> 00:13:38.639
So maybe just pause for a second and let you share that next poll.
00:14:20.720 --> 00:14:21.679
Technology.
00:14:22.399 --> 00:14:35.919
Right, so um ML AI engineering fifty-four percent, and then platform engineering data science, data virtualization, uh data analysis thirty-one percent, and data engineering fifteen percent.
00:14:36.080 --> 00:14:40.399
So probably no surprise in the ML space, but uh yeah, yeah.
00:14:40.480 --> 00:14:55.840
I think so and and yeah, it's uh it's actually really interesting because I think had I asked that same question, say five or so years ago, maybe five to ten years ago, I don't think platform engineering and ML engineering would even have really featured.
00:14:56.000 --> 00:14:59.279
And I think the biggest gap would have been within data engineering.
00:14:59.360 --> 00:15:11.279
And it's certainly something that I experienced um you know in recruiting for roles, and particularly in my my last role in industry in the GNECO, found it really difficult at a point in time to be able to recruit for data engineers.
00:15:11.360 --> 00:15:14.080
There was just too much demand and not enough supply.
00:15:14.240 --> 00:15:17.519
But what I think is just really interesting here is that that's obviously started to shift now.
00:15:17.600 --> 00:15:28.320
You know, we've started to get that kind of level of supply in the market, people have upskilled, reskilled, etc., within the data engineering space, and now we're seeing the next thing happening within AI engineering.
00:15:28.399 --> 00:15:35.039
So just think it's quite interesting, and it's almost an example of where the market moves with the technology.
00:15:35.200 --> 00:15:43.679
So I think that's just kind of called out really clearly that in actually in another five years' time, we are unlikely to see that same level of skills gap in ML and AI.
00:15:44.320 --> 00:15:44.639
Good.
00:15:44.720 --> 00:15:48.159
So thanks, thanks very much everyone for sharing your thoughts there.
00:15:48.639 --> 00:15:55.279
And as I mentioned, those are kind of the key roles that I would see across a successful data team.
00:15:55.519 --> 00:16:03.759
And actually, some of those roles might be the same person carrying out two roles and maybe smaller organizations with smaller data teams.
00:16:03.840 --> 00:16:15.600
You might have someone doing your engineering and your data science, may have someone else doing your data vis and your data analysis as a BA, but it's typically those kind of skill sets that we would look for across the team in general.
00:16:16.320 --> 00:16:37.279
But then to really make your data team successful, I think it's so important to remember the collaboration across the organization, both within your team itself and within the technology function, and secondly, across the actual organization who you're serving and supporting.
00:16:37.440 --> 00:16:45.039
Now, maybe just touch on an anecdote and taking that first path on collaborating across the technology space itself.
00:16:45.200 --> 00:16:58.000
And it's something actually I really experienced back in my time in Agreco and understanding the value of collaborating, of sharing ideas and of kind of building joint teams across the tech function.
00:16:58.960 --> 00:17:08.400
So while I was at Agreco, we implemented a modern data platform, so a single cloud-based platform to support all of our data work across the organization.
00:17:08.559 --> 00:17:14.880
That included sending messages from our AI and ML models to our in-house applications.
00:17:15.039 --> 00:17:24.640
It included producing Power BI or standard tabular extract data for our business, and it also enabled them to self-serve.
00:17:24.720 --> 00:17:32.000
So it really fed kind of all of the data needs that we had within the organization, centralized within this one platform.
00:17:32.720 --> 00:17:39.039
But in all honesty, we we struggled to implement it successfully over the first kind of 18 months.
00:17:39.200 --> 00:17:42.640
We'd aimed to have it in place, and around about 18 months, that was our project plan.
00:17:42.720 --> 00:17:49.839
But as we got to the kind of year point, we realized that actually it wasn't going to be in place, and we learned an awful lot along the way.
00:17:50.079 --> 00:18:11.200
But the real game changer for us and where we turned the program around and really accelerated the pace at which we were delivering is when we embedded one of our software developers within our data team, and we started to apply the principles of the software lifecycle development cycle.
00:18:11.359 --> 00:18:18.960
So, actually, we started to think about how do we think about this data platform almost as a software platform that we are then applying to data.
00:18:19.119 --> 00:18:26.799
And as soon as we started to apply those kind of rules and principles, and it was not overly complex at all, it was thinking about it in a different mindset.
00:18:26.960 --> 00:18:40.640
We really accelerated our delivery, we reduced down the number of defects, we got our code through much quicker, the quality was better, and it was just from thinking about blending all of our technology knowledge together in order to be able to get to that answer quickly.
00:18:40.799 --> 00:18:51.359
And I think that leads into what I was talking about earlier on with platform engineering and really thinking about the platform itself, its scalability, um the cost, being able to control that, the speed, etc.
00:18:51.680 --> 00:18:55.839
And really starting to collaborate across other parts of technology can help.
00:18:55.920 --> 00:19:02.880
And in really large organizations, you'll have a lot of opportunity to do that and by kind of tapping into some of the some of the apps teams, too.
00:19:02.960 --> 00:19:13.680
So that's just kind of one thing, and one thing that can really help also to to kind of upskill your data team and some of those tools, technologies, and processes too.
00:19:14.240 --> 00:19:19.119
And then the second thing I wanted to touch on in collaboration is just the collaboration with the business.
00:19:19.440 --> 00:19:28.319
So I'm not going to talk too much today about data strategy and how you make sure you're adding value and make that um make that impactful across the organization.
00:19:28.480 --> 00:19:39.279
Um but collaboration is absolutely key to make sure that first of all you understand the context of the data that you're working with, whether you're solving a simple or a complex problem.
00:19:39.599 --> 00:19:53.279
And then, second of all, that when you do develop a report, a dashboard, a model, that then the people within the organization utilize your output within their decision making because that is how we add value.
00:19:53.440 --> 00:20:02.400
So, although we're talking today mostly about data skills and how we make sure we're flexible, adaptable, and the collaboration across the organization is absolutely critical.
00:20:02.559 --> 00:20:15.680
And in every single one of these particular roles, communication and people skills are key because we need to make sure that whatever it is that we are creating is picked up, it's used, and it adds value to the organization.
00:20:17.440 --> 00:20:17.759
Okay.
00:20:18.640 --> 00:20:23.039
So I'm going to move on now a little bit to kind of recruiting and retaining talent.
00:20:23.200 --> 00:20:28.000
So when we asked at the poll at the start, you know, how do people find kind of trying to recruit for the right skills?
00:20:28.160 --> 00:20:29.279
What does that look like?
00:20:29.359 --> 00:20:31.039
Um, everyone said that was difficult.
00:20:31.119 --> 00:20:38.079
And I think we we do see that, you know, there's a demand for data and tech skills across the market, and our supply has not caught up generally.
00:20:38.880 --> 00:20:59.279
So, what I wanted to do now was just to talk through some examples of where I've had success and being able to recruit, and then really crucially retain talent as well to make sure that when you get those right people in the door, that actually they're a good fit for the organization, a good fit for the role, and you can then retain them for a reasonable period of time.
00:21:00.240 --> 00:21:03.119
So the first thing I wanted to touch on were apprentices.
00:21:03.599 --> 00:21:14.720
We're very lucky in the UK that we are offered an apprenticeship levy and in UK organisations, which means that the cost of taking apprentices on to organisations is very, very low.
00:21:15.200 --> 00:21:24.960
We had an apprenticeship programme specifically within the software development team within Agreco that we had absolutely huge success from.
00:21:25.200 --> 00:21:40.559
We brought in two or three apprentices per year, straight from school, and who joined our team and then they kind of cycled through a number of different roles within the software, the software dev team, and they added an absolutely huge amount of value.
00:21:40.799 --> 00:21:48.400
So I know sometimes people are apprehensive and want to recruit recruit experienced hires, they want people just to hit the ground running and to get go.
00:21:48.559 --> 00:21:59.759
But I promise you, investing in that kind of pyramid structure where you have a lot of resources at that junior level that you're training up, that you're upskilling, um, can add absolutely huge value to the organisation.
00:22:00.240 --> 00:22:11.279
And also brought some excellent ideas that actually some of our experienced team just hadn't thought about because they were coming onto things fresh phase, you know, brand new and asking kind of the right questions.
00:22:11.440 --> 00:22:16.559
So I would absolutely encourage you to have a think about apprenticeship programmes within your organization.
00:22:16.720 --> 00:22:26.319
And there are also lots of third parties that offer kind of data-specific training programmes through the apprenticeships who can then really support.
00:22:26.400 --> 00:22:35.119
So the apprentices will go off, you know, every couple of months, do some kind of external training with your third party and come back and apply some of that within the organization.
00:22:35.200 --> 00:22:47.440
So it's a really excellent scheme, and I would really encourage thinking about apprentices across data technology dev teams, and they can add a huge amount of value, and we had really positive experience with them.
00:22:48.640 --> 00:22:54.960
Next one I want to talk about is graduates, which is where I think lots of people do and tend to recruit for their teams.
00:22:55.039 --> 00:23:00.640
So, and again, we have done that across all of the roles and in the data teams that I've worked in.
00:23:00.880 --> 00:23:07.680
One thing I would say is that I've not always recruited from technical um degrees.
00:23:07.920 --> 00:23:14.079
So I've recruited people before into technical roles who studied arts degrees, for example.
00:23:14.319 --> 00:23:32.799
And I think that's a really I think that's something to kind of be very open-minded about, particularly when thinking about those data visualization um roles or the kind of data analysis type roles, and where it's kind of light touch on technical skills, but they tend to be very, very strong on communication skills.
00:23:32.880 --> 00:23:42.720
So it actually offers a really nice blended feel to the team and again a different way of thinking, which can only add strength to the overall team and how the team makes decisions.
00:23:42.799 --> 00:23:51.039
So yeah, graduates kind of standard, I'm sure most people use them, but I would encourage you to think about um recruiting from non-technical backgrounds as well as STEM.
00:23:52.559 --> 00:24:05.839
Postgraduate too, so again, thinking about people who have come out from PhDs or MSEs, etc., and bringing that a little bit more, usually deeper technical knowledge within the team.
00:24:06.160 --> 00:24:27.200
Um again, would recruit for those across all of the teams and typically into, as I said, more technical roles and with a particular skill set so that they've studied something like LLMs or something like that, if you want to be able to introduce that into your team, and they can bring their academic knowledge into that business context.
00:24:28.559 --> 00:24:43.519
And then the last one I kind of wanted to touch on, and related to the last poll that we asked, is where have people started to see or started to think about upskilling their team members and across the organization into some of these data roles.
00:24:43.839 --> 00:24:52.880
And the reason I wanted to touch on this is I mentioned previously that in Agreco we really struggled probably seven-ish years ago to recruit data engineers.
00:24:53.039 --> 00:24:55.440
We were implementing this cloud-based platform.
00:24:55.599 --> 00:25:05.759
We knew we needed dedicated data engineering support because our data scientists were doing the data engineering work, but we just could not recruit the people that we wanted.
00:25:06.079 --> 00:25:14.160
So after about six months of trying, um, we decided to instead just upskill the existing people that we had within our team.
00:25:14.319 --> 00:25:23.599
People who had worked as DBAs before, who'd worked as BI developers, who had complementary skill sets and whose roles were naturally evolving.
00:25:23.680 --> 00:25:28.079
We were moving away from you from using on-prem into using cloud.
00:25:28.160 --> 00:25:32.720
And so they wanted to transfer those skills to upskill and to give them further opportunity.
00:25:33.279 --> 00:25:44.720
And so, what we did instead is we essentially built a data engineering team with no experienced data engineers, but we had the opportunity to kind of upskill them as we implemented that cloud-based platform.
00:25:44.880 --> 00:25:49.200
And I'll talk a wee bit about how we did that in a minute when I talk about learning.
00:25:49.519 --> 00:26:03.920
Um, and I think the real benefit for us in doing that is that these people were already very technically competent, and they understood our in-house systems and the data and how it worked, and they understood the context of the data too.
00:26:04.079 --> 00:26:09.839
So, really, what we were doing is just laying on an additional technical skill for them to be able to use.
00:26:10.079 --> 00:26:21.759
And I think that's just a really good kind of tangible example of the world is changing, technology is changing, you know, the tools and tech we have access to are becoming more and more sophisticated.
00:26:21.839 --> 00:26:24.240
So, how do we just make sure that we keep up with that?
00:26:24.480 --> 00:26:35.359
Just iteratively adding on skills and competencies as we go and to make sure we're kind of getting best benefit and that people are kind of evolving as the as the tools and the tech do as well.
00:26:37.200 --> 00:26:49.920
And then the last thing I wanted to touch on in recruiting and retaining talent is just being able to tap in and access government or vendor-funded programs to support and enhance innovation.
00:26:50.319 --> 00:26:55.039
And I just touched on a couple of examples of again where we did that within a GRECO.
00:26:55.599 --> 00:27:05.839
So many of you in Scotland will be familiar with the Data Lab and who support a number of students to be able to study data science and related topics across Scotland.
00:27:06.240 --> 00:27:18.000
And we supported a number of the data science graduates in their MSc programmes, and then went on to recruit some of those graduates after the MSc programmes within the team.
00:27:18.240 --> 00:27:24.319
And that was a really good way for us to be able to, first of all, test out some new and innovative ideas.
00:27:24.400 --> 00:27:41.359
We had over a six to eight week period of them carrying out their project with us, they got a bit of a flavour of what the organization was like, and so it was a really easy transition then when they completed their MSc and to offer them a role within the team, and they came and we had a great success of a number of people coming to join us.
00:27:41.440 --> 00:27:56.000
So it's a really good way of being able to kind of tap into different talent and also through those kind of MSc projects, six to eight weeks to test out some of the innovative ideas you have in the team, but without putting at risk any of the other operational delivery.
00:27:57.279 --> 00:28:01.599
And then the last thing I wanted to touch on was the was working with Innovate UK.
00:28:01.759 --> 00:28:06.240
So there's a number of government bodies that offer funding for innovation.
00:28:06.319 --> 00:28:08.559
So Innovate UK is one of them.
00:28:08.799 --> 00:28:22.799
You can apply for funding over a period of time, and if that is granted, you then work collaboratively with academia and with someone who is sponsored by Innovate UK to tackle an innovative problem that you have within the organization.
00:28:23.039 --> 00:28:32.480
And that again is just a really fantastic way of being able to, at very low risk and low cost, test out some innovative ideas that you have within the team.
00:28:32.640 --> 00:28:38.799
Maybe you've never tried to implement a GenTic AI before and you want to try that, you want to know what it might look like.
00:28:38.960 --> 00:28:54.160
So you come up with a kind of two or three year programme where you're going to do that, you select a university you're going to work with, it gives you that collaboration between public sector, private sector, and academia, and essentially delivers great value at very low risk for the organization.
00:28:54.319 --> 00:29:07.359
So if you're not doing so already, I would absolutely encourage you to have a think about any of those schemes and think about how you can utilize them, as I say, mostly for low risk and low cost to be able to test out some of new innovative ideas.
00:29:09.279 --> 00:29:26.799
Conscious of time, I'm going to say a couple more slides now in terms of keeping skills relevant and when you've got that team in place, what do you do to be able to maintain um to keep them within the team and for them to be able to maintain their skill set and to grow that through their time with you?
00:29:27.599 --> 00:29:50.160
And it's something we used to talk about quite a lot, actually, across all of the kind of data teams that I've worked in, where I think particularly at more junior grades, they were very keen to make sure that there was dedicated time for learning, what the learning program would be, how they were going to grow in the role, how they were going to develop, and wanted kind of very concrete answers to some of that.
00:29:50.240 --> 00:29:53.920
So I'm going to talk through how I kind of typically approach that.
00:29:54.160 --> 00:30:01.680
And often a lot of the ask was can we have structured time and dedicated time every week to be able to dedicate to learning?
00:30:01.839 --> 00:30:07.119
And so I'm quite keen to understand from this group on the call, um, is that something that you typically offer?
00:30:07.359 --> 00:30:10.240
Is that difficult to find time for within the organization?
00:30:10.319 --> 00:30:13.119
Um, so we can pop that next pull up, that would be great.
00:30:39.039 --> 00:30:42.880
So eighty eighty-seven percent know we struggle with that question.
00:30:43.599 --> 00:30:46.079
And and I think that's that's quite typical, isn't it?
00:30:46.160 --> 00:30:50.960
Like in organizations, there's always demand, there's always a deadline, there's always something that's really urgent that comes up.
00:30:51.119 --> 00:31:00.720
And it's always really easy to drop that thing which is kind of additive, it's learning, and then it might not be directly relevant to the deadline or the thing you're trying to deliver next week.
00:31:00.960 --> 00:31:11.920
And it definitely has been a challenge in teams that I've worked across and people just really wanting that kind of dedicated time and it being difficult to try and always, you know, ring fence those couple of hours a week or whatever it might be.
00:31:12.160 --> 00:31:27.039
Um so what I started to do was to encourage teams to think about learning in four different ways and to help them to reflect on how and when they are learning, particularly when that specific or bespoke time that was ring fenced just doesn't happen.
00:31:27.440 --> 00:31:31.359
So the first one I would talk about is on the job learning.
00:31:31.519 --> 00:31:35.119
So I've talked quite a lot about how data and tech world has evolved.
00:31:35.279 --> 00:31:37.920
There is no option but to learn on the job.
00:31:38.079 --> 00:31:47.440
You are either learning a new skill, you might be learning, you might already use Python, but actually you're going to use a new Python library for the new pro problem you're trying to solve.
00:31:47.759 --> 00:31:54.400
You might be working on a new business problem with new data and a new context that you haven't previously understood.
00:31:54.559 --> 00:32:05.599
And I think helping your teams to recognize that that is also learning, helping them to reflect through their personal development plans on what they know now versus what they knew six months ago.
00:32:05.920 --> 00:32:13.359
I think just helping people to understand that when we work in an industry like data and technology, you have to learn just a standard.
00:32:13.440 --> 00:32:18.160
So on-the-job learning absolutely counts and absolutely happens across all data teams.
00:32:18.240 --> 00:32:19.839
I have no doubt of that.
00:32:21.359 --> 00:32:26.640
The second one I think is the one that we're often asked about, which is like independent, or I've said here, or group study.
00:32:26.880 --> 00:32:31.519
So where possible, having weekly focus time to be able to meet some goals.
00:32:31.680 --> 00:32:42.400
And I think in particular, it's it's always helpful to target that against wanting to get, for example, another Microsoft certification or wanting to complete a course on Udemy or Coursera.
00:32:42.720 --> 00:32:50.960
So wherever I've granted that kind of weekly focus time to say, you know, there's two hours on a Friday afternoon, you know, spend it like this, it's with a target in mind.
00:32:51.119 --> 00:32:58.400
It's within six weeks' time, we want to have completed this course, we want to have completed this cert, and we're looking for an outcome at the end of it.
00:32:58.480 --> 00:33:00.880
So it's a really targeted and focused time.
00:33:01.519 --> 00:33:06.559
The other thing is that outcome should obviously in some way be related to the role that they're doing.
00:33:06.640 --> 00:33:15.920
So if someone in the team came to me and said, We used to use Azure in my previous role, I want to go and do this course in AWS, the answer would likely be that's not particularly relevant.
00:33:16.000 --> 00:33:22.559
Um so it should be focused on something that they're working on or something that will be of benefit in the near future.
00:33:23.440 --> 00:33:29.200
The other couple of things we that I've kind of explored in the past which have worked quite well are regular tutorials.
00:33:29.359 --> 00:33:35.759
So typically bringing people in and across your data and tech teams come from different backgrounds, might have studied different subjects.
00:33:35.920 --> 00:33:48.960
So getting people to carry out a tutorial once every two weeks, once every four weeks on something that they are an expert in, and that might actually be presenting back the output of the model they're currently working on within the business context.
00:33:49.119 --> 00:34:00.880
But just setting it up like a tutorial and really framing it like that because it's a learning experience really helps people to kind of link that learning with their role and kind of feel like they're they're still developing.
00:34:01.599 --> 00:34:03.599
Then the last thing is kind of hackathons.
00:34:03.680 --> 00:34:07.200
So I've included that within like groups, group study, if you like, as well.
00:34:07.359 --> 00:34:20.159
Um, and actually I've run a number of hackathons with peer groups, so there might be other organizations who are similar to yours with similar teams, tools, and tech, and getting together, setting up a problem, and then letting people go on and solve that.
00:34:20.239 --> 00:34:33.440
And so taking one day every six months or so and getting together to do that, and again, that just kind of really helps, I think, with sometimes with team building and with upskilling and really helping people to promote additional learning.
00:34:34.880 --> 00:34:37.440
Third thing in terms of learning is classroom-based.
00:34:37.599 --> 00:34:47.840
So I've typically employed that when we have implemented a new tool or technology that we need to learn about, and we want some dedicated and bespoke time on that from our vendors.
00:34:48.320 --> 00:34:59.599
So those could be through people like you know AWS, Google, Microsoft, Databricks, etc., dedicated sessions on their tool and tech and how to get the most out of it.
00:34:59.920 --> 00:35:03.599
Or the classroom-based learning could be on people skills.
00:35:03.679 --> 00:35:11.039
So we've got the output of our model, how do we influence the business, and really thinking about how to bring people in to support in that domain too.
00:35:11.199 --> 00:35:15.599
So another type of kind of learning that people can have access to.
00:35:16.719 --> 00:35:20.320
And then the last thing, um, conferences and peer connections.
00:35:20.400 --> 00:35:36.960
Um and I would always encourage people at every single stage of their career to keep up with their peers that they met at any stage and to keep in touch because you will always learn, be able to bounce ideas off each other and learn something new from what it is that they're doing.
00:35:37.199 --> 00:35:39.519
There are lots of local meetups across Scotland.
00:35:39.679 --> 00:35:42.480
Data Scotland is coming up, I think, again in September.
00:35:42.559 --> 00:35:50.239
Um, international events like Big Data London, um, Data AI Summit, etc., and lots of vendor-led forums.
00:35:50.320 --> 00:35:58.880
Um either recruiters tend to run lots of kind of roundtables, um dinners, meetups, etc., as do software providers.
00:35:59.039 --> 00:36:09.280
So I would just encourage people to think about what they have access to, either remotely, webinars like this one, and or in person, and where there are meetups that happen regularly.
00:36:09.679 --> 00:36:16.320
So yeah, I would just kind of I think that's always a challenge trying to carve out time for learning to keep those skills up to date.
00:36:16.480 --> 00:36:26.079
But those are the four ways that I would really think about learning reskilling and upskilling that can absolutely be embedded with relative ease within your day-to-day job.
00:36:27.119 --> 00:36:28.880
Conscious of time, last slide.
00:36:29.039 --> 00:36:40.800
Um, to finish on, I just wanted to pop on some helpful sources for leaders, and just in my experience, where I've really found it beneficial and to kind of keep up with the market and what's what's going on.
00:36:41.039 --> 00:36:43.280
The first, again, as I said, networking with peers.
00:36:43.360 --> 00:36:58.239
There are lots of senior data groups and roundtables, um, either that work remotely or in person, and those are usually invaluable for being able to share ideas, share what went well, what didn't go well, and learnings.
00:36:58.320 --> 00:37:05.679
Um so, yeah, that's probably my number one um tip is to always keep up with peers and what they are doing.
00:37:05.760 --> 00:37:09.440
Um it really is invaluable to learn from people who are doing the same thing as you.
00:37:10.159 --> 00:37:11.679
Maximising vendor support.
00:37:11.760 --> 00:37:19.199
Um, vendors often have some budget to be able to help you to do some of the things that you want to do, um, either in small projects or otherwise.
00:37:19.360 --> 00:37:21.440
Um so please do lean on vendors.
00:37:21.519 --> 00:37:27.840
I think there are lots of organizations who don't probably do that enough, and they're always quite keen to offer help and support.
00:37:27.920 --> 00:37:39.119
They'll often try to sell you something too, and but usually you can get good learning from them, and also it helps you to keep up with what's coming up next in the market and what other tools or tech might be helpful for your team.
00:37:39.760 --> 00:37:58.079
I plopped in your consultancy support, and I think that one for me is is quite an important one, particularly when trying something new, because consultancies have the benefit that they have already largely done the thing you're trying to do with other organizations and they've seen where it went well or where it didn't.
00:37:58.239 --> 00:38:16.320
So whether you engage with you know a third party to implement something, to get some advisory support, or even just to kind of keep up with what's going on, I would just always kind of keep some kind of connection to consultancies, particularly that operate in the data space, and to kind of keep ahead of where the market's at and what opportunities there might be.
00:38:16.719 --> 00:38:25.679
And then lastly, linked to that, and something I found hugely helpful in my time in Agreco was utilising the Gartner GTP license for technical professionals.
00:38:25.840 --> 00:38:30.480
And for me, that was really helpful, particularly when we were implementing the modern data platform.
00:38:30.639 --> 00:38:36.239
We knew what we wanted to do, but we were pretty agnostic about some of the tools and tech that we wanted to use.
00:38:36.480 --> 00:38:39.679
You can't ask the vendors those questions because their tool is always the best.
00:38:39.760 --> 00:38:47.199
Um so really kind of leaned on Gartner when we had short lists of vendors and saying, here's the things we're considering for these different um functionalities.
00:38:47.360 --> 00:38:48.000
What do you think?
00:38:48.159 --> 00:38:49.199
What would your advice be?
00:38:49.360 --> 00:38:50.880
What's going to work best for us or not?
00:38:50.960 --> 00:38:52.800
And we just got some really good sound advice.
00:38:52.880 --> 00:39:00.079
So for me, that was just really helpful, particularly when making big technology decisions to get a kind of software diagnostic view.
00:39:01.360 --> 00:39:04.480
So um that's been that's all I wanted to talk through today.
00:39:04.559 --> 00:39:16.000
I hope that that's been helpful and just have a talk through current data teams, skills, tools, technologies, and what you can do to maintain that high-performing um data team across the organization.
00:39:17.039 --> 00:39:17.360
Great.
00:39:17.519 --> 00:39:19.039
Thank you so much, Elizabeth.
00:39:19.360 --> 00:39:20.159
Thank you for that.
00:39:20.480 --> 00:39:23.519
Lots of uh hints and tips we can all take back to our visitors.
00:39:23.679 --> 00:39:24.559
I really appreciate it.
00:39:24.719 --> 00:39:27.760
I am cautious of time, so I'm just gonna go straight into questions.
00:39:27.920 --> 00:39:33.280
Um there is a couple of questions with regards to AI and the replacement of roles.
00:39:33.440 --> 00:39:40.559
So I'm talking specifically within kind of data teams and contact with adoption of AI within lower level roles.
00:39:40.719 --> 00:39:45.440
I know you spoke about uh kind of entry-level opportunities and new roles for graduates.
00:39:45.519 --> 00:39:47.519
How do you think AI is going to impact that?
00:39:48.480 --> 00:39:50.320
Yeah, no, I think it's actually interesting.
00:39:50.559 --> 00:39:58.400
I think for me generally, my view on AI is always this it's going to enhance and augment human decision making.
00:39:58.639 --> 00:40:02.000
So I don't think we're going to see AI take over everything.
00:40:02.159 --> 00:40:08.559
I do think we'll see the replacement of some of those kind of, you know, roles which can be automated.
00:40:08.639 --> 00:40:17.679
Um and I think we're seeing that already sometimes, you know, we're no longer right, and we've seen this for years, we're no longer coding out particular statistical models from scratch.
00:40:17.760 --> 00:40:19.199
You know, we pick up a Python library.
00:40:19.280 --> 00:40:20.800
We've been doing that for a very long time.
00:40:20.960 --> 00:40:24.079
So I think we'll start to see those kind of iterative things happening.
00:40:24.320 --> 00:40:33.920
Um I think yes, it has the there is, you know, the risk that it affects those kind of more junior roles because you automate some of those things.
00:40:34.159 --> 00:40:42.639
But I also think that organizations should be savvy in understanding that critical thinking is still a really important skill to have.
00:40:42.800 --> 00:40:51.920
And I think if we automate too much and too early, then people don't really understand all of the things that they are automating and can't really make kind of judgments based on that.
00:40:52.000 --> 00:40:57.360
Um, so I think we'll still see the opportunity for people to be able to do some of those roles.
00:40:57.440 --> 00:41:03.039
I don't think everything will be automated, and I still think we'll just still see a demand for any of the data and tech roles.
00:41:03.199 --> 00:41:05.760
Like I just don't see it reducing kind of over time.
00:41:06.320 --> 00:41:06.639
Okay.
00:41:06.800 --> 00:41:08.320
Okay, no, thanks for that.
00:41:08.480 --> 00:41:14.159
Um, I think following on from that, you've kind of answered that question, but a little bit more on specifically the future.
00:41:14.239 --> 00:41:22.320
So within data and AI, where do you see the kind of most in-demand skill sets then within you know those areas moving forward?
00:41:22.880 --> 00:41:25.360
Yeah, so I think we we touched on it earlier on.
00:41:25.519 --> 00:41:31.519
I think the the demand and the shortage we're seeing just now is that kind of platform engineering and ML AI engineering space.
00:41:31.679 --> 00:41:35.039
I just don't think that we're quite caught up yet with the supply on that.
00:41:35.280 --> 00:41:43.199
I think there's a huge opportunity for people to upskill who are already working in data to be able to operate and start to move within that space.
00:41:43.519 --> 00:41:57.760
But the other thing I think that we absolutely can see, and I think, and I touched on it earlier too, is that people need to have the right kind of people and soft skills to be able to actually get the benefit of these AI models.
00:41:57.920 --> 00:42:03.760
People are always a bit frightened of what might be a black box, like how should we implement it, what is that going to mean.
00:42:03.920 --> 00:42:17.199
So I think really being able to explain what a model is doing under the hood at a very high level, how it can be implemented, what that value might be is really, really key to making sure you have a really kind of successful data team.
00:42:17.280 --> 00:42:24.960
So I think the two shortages for me on the tech side, look at that kind of ML, AI engineering, MLOps, how do you make sure your model's not going to drift?
00:42:25.039 --> 00:42:26.719
How do you make sure you can scale it in the right way?
00:42:26.880 --> 00:42:30.079
How do you make sure costs just don't become you know out of control?
00:42:30.239 --> 00:42:40.239
Um and then on the other side, actually, that kind of influencing, continuing to kind of promote the use of AI and help people to understand it and not be frightened of it too.
00:42:40.320 --> 00:42:41.440
I think there's those two elements.
00:42:41.920 --> 00:42:42.159
Okay.
00:42:42.239 --> 00:42:44.960
No, that's a good answer and uh lots to take away there.
00:42:45.039 --> 00:42:45.679
So thanks for that.
00:42:45.840 --> 00:42:47.599
I think we've got time for one more question.
00:42:47.760 --> 00:43:00.320
So somebody's um asked, how would you advise someone going about challenging or influencing instances where data is produced but potentially being ignored, misused, misinterpreted by management specifically?
00:43:00.800 --> 00:43:01.440
Yeah.
00:43:02.239 --> 00:43:06.559
Um I think it's a good question, and I think it's something that's just very common.
00:43:06.719 --> 00:43:11.119
Um, people often torture the data to get to the answer that they want to see.
00:43:11.280 --> 00:43:13.920
And for me, that's really just a question of leadership.
00:43:14.079 --> 00:43:19.119
Um, so there needs to be the right leadership in place to be able to promote the right use.
00:43:19.199 --> 00:43:32.079
And again, that comes back to the influencing, the tools, the tech, being able to explain what it is that they are doing so that people understand what's happening, and then they take the answer as being the accepted answer and not just something they want to torture to get to the right thing.
00:43:32.159 --> 00:43:38.800
So for me, that's just all about leadership and making sure that the person who understands the model has a seat at the table and is listened to.
00:43:39.280 --> 00:43:39.599
Okay.
00:43:40.000 --> 00:43:40.400
Great.
00:43:40.559 --> 00:43:41.840
Thank you, Elizabeth.
00:43:42.000 --> 00:43:44.159
Um I'm afraid to say that time is up.
00:43:44.239 --> 00:43:47.599
So I just wanted to say thank you to everyone for joining the call today.
00:43:47.760 --> 00:43:50.880
Massive thanks to Elizabeth for delivering that presentation.
00:43:51.039 --> 00:43:55.119
Lots to take away um that we can all use into our businesses moving forward.
00:43:55.199 --> 00:43:56.480
So I really do appreciate it.
00:43:56.719 --> 00:43:57.039
Thank you.
00:43:57.519 --> 00:44:08.800
Um just finally from me, I We'll be hosting our final Dayton AI webinar um in a couple of weeks with Callum Sinclair from Burness Paul talking everything AI and legal.
00:44:08.880 --> 00:44:11.360
So look forward to seeing you on the call then.
00:44:11.519 --> 00:44:11.760
Okay.
00:44:11.920 --> 00:44:12.880
Have a great day, guys.
00:44:12.960 --> 00:44:13.280
Thank you.
00:44:13.360 --> 00:44:13.920
Bye bye.
00:44:14.079 --> 00:44:14.639
Thanks, Al.
