hitGlobal_edit_podcast_1
[00:00:00] Speaker: Welcome to the next episode of the Humanizing IT podcast, We Need to Talk About IT. And yes, we do. On that note, my next guest is someone that I admire. Uh, I met this guest probably about six months ago at a conference in the United Kingdom, at an international Woman's Day, actually, for IT. Uh, this guest was the emcee.
[00:00:22] I was super, super impressed with her ability to engage the room, and also her ability to really, uh, explain and help what's going on in the IT industry today. Our guest works for one of the largest corporations in the UK, and I'm proud to introduce Valerie Wilson.
[00:00:40] Speaker 2: Thanks very much. I'm delighted to be here.
[00:00:42] Thank you very much for asking me, Katrina. Mm-hmm.
[00:00:44] Speaker: It's, as I've said, a absolute pleasure, and you are the, as, the first woman we have had.
[00:00:49] Speaker 2: Very... I feel very honored- ... on behalf of all women. Hopefully I'll do it justice.
[00:00:54] Speaker: Yeah, but-
[00:00:55] Speaker 2: Hopefully, yeah ...
[00:00:55] Speaker: uh, it's interesting, and, and as I said, we, we like to take a bit informal, these podcasts.
[00:01:00] I did say to you, Val, that I have been, um, sort of pulled up to say, "Hey, Kat, how come you and Wes only have men?" Mm. And I do ask women, and, and most of them don't respond, or, and I'm not sure why that is. But when I ask a man, they jump at it. And so yeah, it's, it's, it's, you know, it's an interesting-
[00:01:21] Speaker 2: I don't know why that w- I don't know why that is.
[00:01:21] Mm. I, like I... I've done a few podcasts. Not loads, but I always, I al- I, I might be wrong, but I, I love, I love a chat, you know? I love a good chat, you know?
[00:01:30] Speaker: Yeah, no.
[00:01:31] Speaker 2: Work and what's happening in the industry- Absolutely. Yeah ... the chance to, to share with what's going on. Mm, mm. So suppose everyone's different, but yeah, I can, I can imagine some people, but maybe more so females, are a little bit reluctant- Mm
[00:01:43] do you know? 'Cause it can be a bit daunting, let's be honest- Yeah ... you know, that way. Yeah, mm. You know, maybe you're going on and you're effectively just talking live about whatever it is that, you know, the person wants to talk about. So I can sorta get it. Yeah. But yeah.
[00:01:54] Speaker: No,
[00:01:55] Speaker 2: yeah. Like I said, it's down to the, the confidence thing,
[00:01:57] Speaker: you
[00:01:57] Speaker 2: know?
[00:01:57] It...
[00:01:57] Speaker: I, I agree. Yeah. Mm-hmm. So thank you for, for being the first. Yeah. Now Val, your role is Manage Services and AI Ops Program Director for British Telecom. Have I got that right? Mm-hmm. Fantastic.
[00:02:11] Speaker 2: Do you know what? I'm only, I'm o- I'm not even sure myself of my job title, but if that's what it says on LinkedIn-
[00:02:16] it must be true, isn't that right?
[00:02:17] Speaker: That's, that's what it says on LinkedIn. And- Yeah ... uh, if you could tell the audience a bit about you. Uh, uh, I said that, you know, the things you do personally are... You're, you're such- Yeah ... a busy lady, and professionally. Yeah. Just, uh, tell us a little bit about you.
[00:02:31] Speaker 2: Yeah. So, so I'm Valerie Wilson, for those of you who don't know me.
[00:02:34] Um, I, I live in Northern Ireland, in Belfast, so you can probably tell by the accent. Um, I think on a, on a personal level, first of all, I am married to Mark. I've got three daughters, Hannah, Jessica, and Amy. Um, so they're 16, 14, and 11. So some people say good ages and then some people say bad ages, but yeah, I, I quite like that age.
[00:02:52] There's less having to be there holding their hands, but yet more importantly emotionally probably, um, for girls at that age. So that's the kids. I, as you say, I, I pack a lot in to, um, I, I do quite a bit of sport and fitness. Um, this season again I've started to play football again, as well as coach football.
[00:03:11] Um, so I've always played football over my, my lifetime, and I sort of went into retirement to coach, and then the club I'm at started up a senior ladies and they're like, "Val, we need you to come back and play." So I've dragged back into actually playing football and in mid-40s, it's not probably ideal, but there you go.
[00:03:26] We're still keeping up with the kids, shall we say. But yeah, so I play a lot of football. Um, I also run a youth club, um, for our local church, and um, I sit on the board of governors, and I also sit on the board for ITSMF UK as well, and chair the women's network. So I suppose from a personal point of view, I do, as you say, I, I'm one of these people that if there's something needs done, I'll be like, "Yeah, I'll do it."
[00:03:49] Or if there's a committee needs done, I'll end up running it and I need to stop because my, my days are becoming jam-packed. And then as you say there, I have, um, a really exciting job at work at the minute as well. So I've worked for BT for 25 years, my whole career, um, in BT. Um, so I've done a lot of different jobs over that time.
[00:04:07] Probably for the last 10 years, focused more on, um, IT service management, managed services. Mainly across my career has been leadership roles, but this most recent one, which is only since January this year, is all about implementing AI into IT service management and how can we do things better, quicker, faster.
[00:04:25] So a lot of buzzwords in there, which means I get a lot of cold calls now than I ever have done in my life, do you know, from, from, from LinkedIn, and I wonder who got my number, um, all the time. But, uh, yeah, it's a r- I'm really fortunate to have this role, um, I have to be honest, and it is exciting, but it is, it is tricky, but it's good.
[00:04:42] Mm. It's good. It's all good.
[00:04:43] Speaker: Mm, 'cause yeah, AI, AI, AI, that is-
[00:04:47] Speaker 2: I know.
[00:04:48] Speaker: Tell me about it ... everything we hear. And, and it's interesting your role, but if you could explain, uh, 'cause I think it's a, a new, I, I just did an a, a podcast with Akshay, and, uh, we were talking about how AI is impacting roles in IT service management.
[00:05:05] So what is AIOps?
[00:05:09] Speaker 2: Yeah. So really if, if you break it down and ke- try and keep it really simple with AIOps, it's using big data. And when I say big data, it's big, right? It's, it's more data than you can really use in terms of every day. So what you're using all types of feed from a telemetric point of view, and then you're running machine learning over the top of that effectively to do things quicker, better, faster.
[00:05:31] Yeah. So it's that combination of big data, machine learning, and then your automation and your AI over the top of that. So the way I, I try and keep it simple and go, you think of all of your feeds that you have today that gives you information, just to keep it really simple. You bring them all into one area, and then you run effectively your machine learning over the top of that, and then you use the different types of AI to really try and pull to the top what are the best processes, the best, most repeatable tasks based on the history that we've seen, and indeed what we think is gonna happen in the future, and how can we run that quicker and better.
[00:06:04] That's probably, in my opinion, the easiest way to describe it. But yeah. In the operations, it's in that operational area.
[00:06:11] Speaker: But how would that work? 'Cause we, we did talk about this earlier, that you're focusing on AI for IT service management versus-
[00:06:19] Speaker 2: Yeah ...
[00:06:19] Speaker: AI for BT business, which I think is fascinating. Yes.
[00:06:23] Yeah. W- In, in regards to IT service management, what big data do you get? I mean, is it incidents or, um-
[00:06:29] Speaker 2: Yeah. It's, it's across a lot of d- of different things. So as you say, I'm very, at this stage, I'm very specific at managed services. And when we say managed services, what it means, and this can mean different things for different people- Yeah
[00:06:39] but in, in, in my context, it's effectively where do we proactively monitor our customer's infrastructure, right? So BT, it's not really a desk, you know, it's not really let's look at, at the internal IT. We run our platform across thousands of, you know, external customers, um, and it's across any technology.
[00:06:58] So it could be network, could be, um, firewalls, it could be data, it could be voice, it could be, um, server storage, it could be anything at all. So our big data, back to your question, effectively comes from all of our different tools that we use to monitor, um, the performance of the customer's technology. So it, it comes from probably a range of different things.
[00:07:17] And equally because we're BT, we obviously run the core network as well. Mm. So we also take feeds from the core network itself. So we've got feeds from the core network and from the customer's network- Right ... whether across all their technologies. So really then you're getting this huge lake of data that, that we're able to then interrogate using the AI, the machine learning, to try and do things better and quicker and faster.
[00:07:41] So in simple terms- I, I always, when we talk about a managed services, I, I describe it as Big Brother, right? So Big Brother is watching you, i.e., we are monitoring. Uh, so managed service and monitor, they're the two things. So M and M, managed service, monitor, and that's ... And if it's not monitored, then that's just a service, not a managed service.
[00:07:59] Speaker: Right.
[00:07:59] Speaker 2: And then you don't have as much rich data-
[00:08:02] Speaker: Mm ...
[00:08:02] Speaker 2: effectively to be able to run your machine learning, your AI over the top of. But because we run a managed services, and because we've got lots of different tools in our stack, we actually are in a real, I think it's a real advantage to have that information and then use that effectively to do things better and quicker and faster.
[00:08:18] Speaker: So when, when you say that, and again, I'll, I'll go back to service management, and this might be a stupid question- Mm ... but typically-
[00:08:27] Speaker 2: There's no such things. Might, might be a stupid answer, to be honest with you, Katrina.
[00:08:30] Speaker: Who knows? Who knows? And then this is a cynic coming out of me, Val, but typically when an IT organization embark on a, a new -- If we take AI out for a moment, embark on a new tool or a new transformation- Yeah
[00:08:41] we start with the usual suspects: incident, problem, change, and service request. Yes. Is that what you start with?
[00:08:48] Speaker 2: Yes, absolutely.
[00:08:50] Speaker: Oh, do
[00:08:50] Speaker 2: you? Absolutely. It i- Yeah, it is. I know it's not a stupid question at all. I, I, I tell you why.
[00:08:55] Speaker: I was hoping for a different answer.
[00:08:57] Speaker 2: No, I, I'm sorry. It's not a different answer.
[00:09:00] Maybe that's the wrong answer, but it's not. I, I'll tell you why. Um, whenever we've looked at our business case around doing it, you know, we have to look at where the bulk of the effort is, right?
[00:09:10] Speaker: Okay.
[00:09:11] Speaker 2: So whenever we looked at where the bulk of the effort is being spent in our IT service management world, as you say, it's incidents, it's service requests, and it's change, right?
[00:09:21] Now, ultimately, let's say whenever you, you've finished that bit and you wrap a bow on it and it's all nicely done, which dear knows when that will be, of course you would start to look at those wider ones. Do you know, you would start to look at capacity management, service continuity management, asset management, right?
[00:09:35] You would look at all those ones, but the first ones, and it may just be the way that we're set up and our effort, but the first ones were the most effort. And also from a customer perspective, you probably want to improve first is your incidents, your SSRs, your change. Do, do you know? There you go. It's not, not rocket science.
[00:09:52] Oh,
[00:09:52] Speaker: dear, oh, dear. And here I was thinking that AI's gonna change how we do things. But why don't you, um, start with problem management?
[00:10:02] Speaker 2: Yeah. So, it, it's, and there is obviously inferred benefits to problem management, right, when you look at incident. But the reality is, a- and, and again, I think if you had an endless pit of money-
[00:10:13] Speaker: Yeah
[00:10:13] Speaker 2: Katrina, right? You have to be able to prove your return on investment quite quickly. And I don't know if it's just a BT thing, but our rules are pretty stringent about CapEx and investment. So you have to be able to demonstrate the payback as well as, of course, do you know the benefits for the customer- Yeah
[00:10:31] do you know, and everything else. And therefore, the effort is involved. So our problem management team is relatively small in comparison to our incident management team, and that's probably as a result. Now, don't get me wrong, there are definitely some what we call quick wins or out of the box areas that would impact problem.
[00:10:48] So we're not saying we're not touching problem, but it's probably just not a radical, do, do you know what I mean? Let's start with problem first and not look at anything else. I think that would definitely be wrong. But to give you an example, um, in incident management, one of the, one of the use cases we're looking at is around auto of closure codes.
[00:11:02] So how do you get the closure codes right every time? Now, ultimately, if you do, if you get that right through, um, machine learning, right, and as we know, people at the minute when they do their closure codes, you know, you sometimes you get a bit trigger-happy. Right. And you go with whatever's the easiest thing, everybody not quite sure, and that obviously impacts problem management 'cause they're working- Sure
[00:11:19] off data-
[00:11:20] Speaker: Mm.
[00:11:20] Speaker 2: Right ... that's not real. Mm-hmm. Whereas whenever you use AI for incident to get the closure code right, ultimately helps problem management. Similarly with the ticket summarization to know how do you summarize things quicker, that helps problem m- whenever they're looking at their RFOs to know when their PIRs that they're trying to do post investigation around that.
[00:11:39] So, so it's not that we're not doing problem management- But there's elements of incident management- Mm ... that will make problem management a lot slicker around that. Um, and we do have some further use cases, and this is probably another thing around the data. So problem management relies on the data, and with any AI program, your first year really you need to get your data right.
[00:11:58] You know, your structure right, your governance right, your correlation right. Um, and after that you can start to see benefits flowing through into other areas where maybe you can interrogate the data be- better. You know, there's a great tool that we're putting in, which is around, um, how do you get good business insights from the data?
[00:12:14] A bit like how do you use Copilot over all that big data that we're talking about. Mm. Now, that will massively help problem management, but you can't do that-
[00:12:21] Speaker: Yeah, unless you have the data ...
[00:12:22] Speaker 2: your first probably year, year and a half- Mm ... until you get all of that in place- Mm ... and everything right. Mm-hmm. So there's probably gonna be more implied benefits down the line that we maybe aren't even know ourselves yet until we get that- Mm
[00:12:32] that part done. But yeah, that's probably ... I would say it indirectly impacts for a benef- better problem management, but I don't think, I don't think you would ever just go, "Let's look at problem management first," and do it. Do you know? I think you would probably do it the way we're, we're looking to do it.
[00:12:46] Speaker: It, it's a, an interesting view that I'm kind of embarrassed that I really haven't thought of this before. In some ways, um, you need incidents to do better problem management.
[00:12:57] Speaker 2: Well, that's some very interesting incident. There's no, there's no other way. That's a ... Well, let's ... That would be utopia, wouldn't it?
[00:13:03] Or, you know, it would be absolute utopia if there's no incidents.
[00:13:06] Speaker: Well, that's right. I don't know if that ever happens. So next time there's a major incident, um, everyone can thank you, Val, because y- you, you're just saying you need the data.
[00:13:15] Speaker 2: That's exactly right. Exactly.
[00:13:17] Speaker: And, and again, within your, um, experience and obviously the research you're doing with your role, um- What's the hardest part about AI, implementing AI for- Right
[00:13:30] IT service management? Do you implement it? Is that the word, or what, what is the word you say?
[00:13:33] Speaker 2: Yeah. Yeah. Well, yeah, you could say that. Yeah. I mean, 'cause a right or wrong word to say- Yeah ... but probably implementation's the, the right word. Oh, man, th- there's, there's quite a lot, and, uh, t- I, I think at, at different stages of your program, depending on the maturity of the program, there'll be different challenges.
[00:13:47] Um, I think initially the biggest challenge was getting a business case, you know, that worked, you know- Oh ... in terms of getting that payback that we talked about there. It gets death have been difficult. Hmm. Um, I would say as we were progressing through, if you like, this data governance stage, that's another big problem, do you know, is making sure your CMDB, your inventory is good and, and with all of these programs, and you'll hear it over and over again, and now that I'm in it, it is so true, that first year of, of any AIOps program, there's very little, I'm gonna say wow moments, because the data has to be right before you can get, get to the good stuff.
[00:14:24] And a lot of people, 'cause to your earlier point, AI is the big buzzword, everyone wants to know, like I spend most of my time updating, you know, s- more senior people than me about how the program's doing, and there is a impatience, let's say-
[00:14:36] Speaker: Hmm ...
[00:14:36] Speaker 2: do you know, you know, from people who maybe aren't so close to it or don't need to be close to it, whereby they've gone, th- they've given you the check, do you know, and they wanna see the results there and then.
[00:14:47] But that first year, year and a half, there's so much work and effort required to do on your data structure, your CMDB, and making sure that is right. Because if you don't have that correlation right, as I say, all that good, do you know, to be in AI or the, the more sexy technical stuff that people expect and you hear about AI doing, y- you can't get to that.
[00:15:07] So I think there's an element of that first year being tricky to get the data right and the expectation. And the final thing I would say is around adoption, do you know, um, value realization. Um, and again, it's maybe it's the stage of the program we're in now- Hmm ... but whenever you're implementing, you know, some of the, let's call them the quicker wins, you know, you are relying on people, you know, you know, I'll need to hit this button to make sure this summarizes the ticket, or I need to hit this button to do a, you know, to, to craft the email that's gonna go out to keep the customer informed.
[00:15:39] There is an adoption piece there with people, and I think that fear factor kicks in, d- you know, um, around that in terms of I've always done it this way, and how do, how do we do that? And it's not like- It's not like a, you need a training course to do it 'cause it's pretty straightforward. Do you know, you hit this button.
[00:15:55] You know, it's not like you're taking them all- Yeah ... you know, away and you're going, "Right, let's sit in the room and let's explain this," because it's not difficult per se. It's more of a psyche thing, do you know? And, and, and I get it, and even if you think of, like, some of the stuff we're doing is around, you know, how do we get our auto diagnostics, do you know, everything there if you like on a page for people to see.
[00:16:13] There you go, boom. But people do like to run it themselves. People like just to, "I'll just do a wee check here and I'll run it just to make sure in the background," do you know? Well, act- it's there. You don't need to, but, you know, that, it- it's that, it's getting people out of that way of thinking, and I think that will take time.
[00:16:29] Do you know, that's not something that happens overnight whenever you go live. It's something that ramps up, do you know, over time. Uh, and, and that's a tricky, that's a tricky one as well. And- They would be my three.
[00:16:41] Speaker: It's, it's interesting. Um, I'll, I'll... There's lots of points in there, but if I go back to that, um, dirty word called CMDB.
[00:16:50] Speaker 2: Yeah.
[00:16:51] Speaker: Is that important?
[00:16:53] Speaker 2: Oh my word. Yeah. Like, honestly, you, you, you can't underestimate the importance of that.
[00:16:58] Speaker: Isn't that, isn't that interesting?
[00:16:59] Speaker 2: And it depends how ... I wanna say, and it de- like, I don't know how, what other companies are like with their data, but anyone I've ever spoken to says the same thing.
[00:17:06] That no one's data's brilliant, right? Let's be honest. Yeah. And a lot of that, a lot of it is historic. And certainly from a BT perspective, it's an historic problem. It's usually because of a, a full lead-to-cash journey, you know, problem. It's not necessarily something, you know, that you've got wrong, uh, you know, once something goes live.
[00:17:23] It's right from when something is ordered, delivered, that whole process. Yep. You almost need another whole transformation program just getting that smooth, which we have as well, you know, in BT. But it's, it's not an easy thing to fix that, and we've tried to, um, look at, well, how bad do we think it is at the minute?
[00:17:40] And, and it wasn't as bad maybe as what we thought it was. Mm. But you need to work out quite quickly what are the minimum data sets you require to make all your AI run, and therefore how mu- how much are we gapped on? And therefore, how do we get those, how do we get those- Mm ... filled, do you know, that way- Mm
[00:17:56] to make sure it's in the right way. But we've been working very closely with, um, other people who've done this before. Yeah. I think if you were to go do it right on your own without real expertise there- Yeah ... you'd struggle. So to be fair, ServiceNow have given us, um, a lot of their more, I'm gonna say, expert resources- Right
[00:18:13] to know where they're really helping, helping us to drive the value in that, where they've done it before for other people. The same with Accenture. You know, they've done it before for other people. So they're trying to take ... Albeit our ecosystem and our infrastructure is very different, but in terms of the principles, they, they, they've learnt the lessons the hard way.
[00:18:31] So we're hoping, you know, that that's gonna help, and we're not gonna be here in a year's time and go, "Flip, our, our data's still not right, and our CMDB's still not right." Which I think you could find yourself quite easy in that position if you don't really go at that, I suppose, with the reverence that's needed to go, "This is the most important thing that you need to get right to get the benefits down the line, and make sure it's working down the line."
[00:18:54] You know? It is. And I was chatting to-
[00:18:55] Speaker: Mm ...
[00:18:55] Speaker 2: another lady, Siobhan, um, Hobbs, the other week- Yes ... today, and she was doing something. You know Siobhan?
[00:19:00] Speaker: Yes.
[00:19:00] Speaker 2: And it really encouraged me actually, 'cause she says as well- Year one, you're just scraping the surface. You know, if anyone thinks there's any benefits in year one- Yeah
[00:19:07] they're, they're really, really deluded. Yeah. And I think I'm so glad to hear you say that , you know? Yeah. So glad, because you know, it is, it's year one is a
[00:19:16] Speaker: difficult
[00:19:16] Speaker 2: year- Yeah ... you know, to get through on that, on that front. And, um, I, I think to over-perform in year one in the AIOps program is probably the hardest thing you'll ever have-
[00:19:26] Speaker: Mm
[00:19:26] Speaker 2: ever to try and do. Yeah.
[00:19:28] Speaker: But it, it's kinda like saying, um, frameworks like ITIL, they've always had this governance in play for service management, just, just CMDB and your hygiene of tickets. Yeah. It's kinda like saying, "I told you so." Like you, you need that
[00:19:44] Speaker 2: to- Yeah. That's so true It's wise. It's so true. And people, you always hear people saying things like, "Oh, if you're building a house, you know, you need the foundations, and the foundations is CMDB." The amount of analogies you've heard about CMDB- Yeah, yeah ... and houses, and foundations. Yeah, yeah. Blah, blah, blah.
[00:19:54] Yeah. Do you know? And I'm trying, I'm trying to get the furniture in and haven't even built the walls. Do you know, it's, it's thing, it's things like that- Yeah ... that you hear all the time on different areas, and it's so true.
[00:20:03] Speaker: Yeah.
[00:20:04] Speaker 2: Um, it's so true, but I think AIOps, if you're doing it in your business, this will force that.
[00:20:09] It will absolutely force- Yeah ... your hand to do that. Yeah. And I think that's not, not a bad thing. That
[00:20:14] Speaker: thing. No, a- absolutely. That's a good thing. Because, yeah, I mean, that's why we have these, this governance in service management- Mm-hmm ... is to have a, a smooth and incident-free, uh, or incident reduction environment.
[00:20:24] And Val- Yeah ... w- in relation to, um, I guess day-to-day processes and how we know them- Yeah ... today, we do knowledge management, problem management, and we log a ticket. Yeah. We might functionally escalate it, we give it to a resolver group. Yeah. Will that fundamentally change, do you think?
[00:20:43] Speaker 2: Will it fundamentally change?
[00:20:45] I think what will fundamentally change is the effort required to do that, but I don't think ... Is it gonna fundamentally change? I, I don't think it will, right? Look, uh, potentially, as I say, it'll be a lot quicker than what it's done now, today, but ultimately, ultimately those steps are still the right steps.
[00:21:07] You might shortcut a few, do you know, that way? So for example, I'm just trying to work through an example in my head. Can
[00:21:11] Speaker: I, can I give you one?
[00:21:12] Speaker 2: Yeah, go for it. You go for it. Yeah.
[00:21:14] Speaker: So it, it's interesting you say that, and sorry to interrupt, but part of what we do at Humanising IT is to challenge- Mm-hmm ... what we do today.
[00:21:22] Mm-hmm. Yep. Because I think, um, you know, we talk about something called IT service management confirmation bias, and-
[00:21:30] Speaker 2: Yeah ...
[00:21:30] Speaker: we have a, the quiet architecture in IT service management, meaning that we're following best practice. And, and I always give the example- Yeah ... contact the user three times before you, um, close the ticket.
[00:21:42] Yeah. Now, to me, you know, we've never challenged that mostly in, in IT service management. Yeah. We think we're doing the right thing. Um, another example is close the ticket if a workaround has been provided. Um- Yeah ... so that's why I'm asking, and we sort of see within Humanising IT, AI or any improvement really, gives you the opportunity to say, "Hey, stop.
[00:22:04] Is this actually the most efficient-" Yeah ... efficient way to, to do it?" Um, just 'cause we've always done it- Yeah ... doesn't mean it's, it's the right way to do it. And, you know, we quite often say, "Did anybody ask the user if they wanted to be contacted three times?" And the answer's invariably no. So true. So that's what I mean.
[00:22:21] Yeah. Like, you know, are we using- Yeah ... are you using the time to sort of- Humanize processes, um, or just fundamentally, you know-
[00:22:29] Speaker 2: Yeah. It's a really good, good point. So every time we look at a use case, um, uh, in terms of what we're gonna do, we walk through the process, and then we think about, well, what, what process do we want it to look like, right?
[00:22:40] Yeah. And then how do we use AI to improve it? So invariably, you do look at that, but, but what you'll probably find is people are so used to that way of working that it's quite hard to, to come out of that box. Yeah. Do, do, do you know? And what I would say with some of the, the big ITIL processes that we all know and love effectively, they are high level, right?
[00:22:58] They are, as we know, high level, and typically what it gets into is local work instructions, which is if you like- Mm ... the detail, detail below it. Yeah. So an incident, whatever, you've got your five key steps, whatever it is. You know, but underneath that, in terms of that local work, and I think those broadly speaking, the five stop steps at the top- Yeah
[00:23:14] do you know, are, are right. But whenever you get into actually what happens on the ground- Yeah ... how do people use those steps to navigate through it? Mm. That's the bit I think that needs probably to change- Yes. Yeah, yeah ... do you know, more so than your broad, high level- Yeah ... bit at the top. Like, it would be lovely to get to a stage where you never need a CAB, for example, in change.
[00:23:32] Mm. Sh- show that way. Mm. Because everything has been de-risked down to, rather than a complex or normal change- Yeah, yeah ... do you know, to it, it's a standard change. Change, yeah. Because of, because of the way AI will analyze a change. Mm. Do you know, it will be able to, with, with confidence, say what the impact is.
[00:23:50] Mm. With confidence, say, actually, there's a, there's a, a lack of resiliency over here in advance, so we need to make sure that that's looked at. Do you know, we've already done the rollback plan, we've done this, we've done that and the other. You know, it should be how do we get more changes to be standard, do you know- Mm
[00:24:03] therefore not needing- Mm ... to go through the CAB, do you know, to be able to do that. Um, a- and how do we then work that through? So- I think in Verb, and there's definitely gonna be different skills for people required to do all of these things, and I think what you may find is where elements are going to different resolver groups today, they may not have to go there in the future.
[00:24:21] Yeah. Because AI will go, "This is next step. This is what I think we should do next based on the last thousand tickets that look like this- Yeah ... for this customer." Like, you know, "No, no surprise, guys," do you know what I mean? "We're gonna need to do it like this. Therefore, do you want me to do it like this?"
[00:24:37] And the person there will look at it and go, "Yes, I do." Get on and, and do it. Yeah. Whether that be a change to the VLAN or, you know, changing a config or reboot, whatever it happens to be for, for that thing. Um, so you may find you don't have to use those user groups as much, or indeed vice versa. You may find it goes directly to those user groups- Mm
[00:24:55] because it knows based on the history that every time someone at the frontline gets it, that's where it goes to. Yeah. So it cuts out, if you like, the middleman. You know, the middleman could be anywhere in these situations, but ultimately it will, it will get the same outcome, right? Mm. But potentially a lot less steps.
[00:25:12] Yeah. Do, do you know? Certainly a lot less steps from a human touch point of view, I think is probably the way I can see that going down the line. A lot more about people understanding the guardrails and, if you like, checking the work that AI, or checking the next steps that AI is saying. Do you know? So you're always, I think, gonna need a human in the loop somewhere there on, on, particularly with the customers we're working with, which are very much your cr- critical national infrastructure customers.
[00:25:37] Um, so I think that's an important factor for sure.
[00:25:40] Speaker: And do you think the days, it's, um, probably a stupid analogy but I'll go with it. Um, I, I spoke to someone from America recently, and we actually had a meeting, um, and he was in his drive- driverless car. And-
[00:25:55] Speaker 2: Okay ...
[00:25:55] Speaker: you know? And you know, I mean- In the
[00:25:57] Speaker 2: meeting, right.
[00:25:59] Speaker: And-
[00:25:59] Speaker 2: That'd be a bit controlled ...
[00:26:01] Speaker: you know, and, and you know, we, we're fortunate enough to travel, um, the globe and we speak at- Yeah ... conferences and in, in Vegas and, you know, other countries in the world. Yeah. You know, they tr- have true driverless cars. Yeah. And, and to me, I, I think that technology is amazing, you know?
[00:26:14] Yeah. Um, and is service management gonna become like a driverless car, where we don't need people to steer and direct? Uh-
[00:26:25] Speaker 2: And I, I, oh, I, I don't think so. Okay. I honestly don't think so. Interesting. I really don't. I don't. Um, not ... A- and it may just be that it's just too far away at this stage to, to, for it to be a reality.
[00:26:39] I think there's absolutely elements right today where we have humans doing things that we won't need them for, but as a collective and as a whole, when you're running a managed services, the humans are so important, right? And, and that's because customers are so important ultimately at the end of the day, and customers need relationships.
[00:26:58] They need to build trust, and they need to have people like they know are doing that. And like we have, um, we haven't done, I would say, a m- massive launch externally. That's gonna be coming from a BT perspective, but we have spoken to a number of customers to know about what we're doing, how we're doing it, and there's definitely a nervousness, a nervousness and excitement around that.
[00:27:18] But what they will always ask is, "Hold on a minute. How are you gonna prove that, you know, AI knows what it's doing? How are you gonna prove that with confidence you need that?" And what reassures people at this stage, in a way, is that human in the li- human in the loop model. You know, we've got a human there checking, making sure, and I think the governance on the AI is so important.
[00:27:39] You would have to be proven, right, this is 99.99%- Right. Yeah ... accurate at all times, do you know, that way, before you'd remove the human, and there will absolutely be circumstances where you would never, even if it's 100% every time, you would never- Right, the human ... remove that human in the loop because of the types of services, do you know- Mm
[00:27:57] that we are looking after from an AT perspective. Mm. The, the risk and the impact of something going wrong, and like, it's funny because over the years, like all the major incidents, do you know, like, uh, from, uh, Mem- oh, Mem- Memory Time, the majority of times a, a major incident happens because of human error Is mad.
[00:28:14] But yeah, we're talking about AI, and everyone's more freaking- Yeah ... about AI error, and I recognize humans have humans have error as well. Very much so. As a manager, you can see that in so many instances, so- Yeah ... it's quite ironic, do you know, the way, but I think even though they're probably, their error rates is probably same if not worse- Yeah.
[00:28:31] you know, people feel more comforted. Mm. But I think a- as much as anything, it's about how do you interpret what all of this data is telling us.
[00:28:40] Speaker: Mm.
[00:28:40] Speaker 2: And that interpretation, right, come, should be coming from your technical experts in managed services. Yeah. Mm. And playing that back to their counterparts with our customer to have those conversations to make, ultimately make better decisions for their future.
[00:28:54] Mm. And that's, that's what it is. Yeah. You're always gonna need humans there in various forms. A- as I say, I don't think you're gonna need all of the different roles that we have today. Mm. And I think it could be a whole different way of working in it, but during, I don't think, I don't think you'll ever get, "Let's do a managed services as a driverless car."
[00:29:11] Do you know, unless it's a very, very standard service- Yeah ... you're providing. Yeah. Do you know, more the SMD, you know, your, your very small medium business. You're not- Yeah ... getting into your bigger customers. Yeah. Do you know? Like, there's definitely an option for that for the low-end market, you know, where- Mm
[00:29:25] where, you know, a different price point, but when you get into your more complex managed services, I don't think that's gonna ever be the case, to be honest- No ... with it.
[00:29:34] Speaker: And, and to be honest, that's really great to hear, you know, someone like yourself being, actually doing- Yeah ... AI in, in ops, to hear that, that we still always need a human in the loop.
[00:29:44] Um- Yeah ... and it's not always, and it, and you know, good news too, it's not that roles, well, roles will go, but skills will change, and, and that's- Yeah. Absolutely ... again, an opportunity. So yeah. Mm-hmm. Uh, and, and just before we go, I did wanna mention that, um, you know, another reason I was so excited that you s- accepted to appear on the podcast, and I mentioned that we go to Vegas quite often, and the last conference we did was called Humanizing AI.
[00:30:08] Yeah. And, uh, we actually were at full capacity. Now we do training on, on this, and we normally have 20 people in a class, and unfortunately, might be my fault, um, the conference organizers didn't ask how, what's the capacity you can have. And anyway, we had 100. in a class and the demand, and they actually had to cut it off 'cause they didn't have the room but, and we had an...
[00:30:32] I won't mention what the other classes were, but they, they had the 10 and 20 in the class, and we were packed. Yeah. And it's just so interesting that this whole AI, you know, industry, the, the interest that we have, and I think that-
[00:30:45] Speaker 2: Yeah ...
[00:30:45] Speaker: there's not enough information for, for people, so that's why it was, s- really great for you to come on this podcast and actually- Mm-hmm
[00:30:52] what, what is it, and what's it like day to day? And, and for that, I'm, I, I appreciate, and I'm sure our, our listeners do also. Um, and yeah, so on that note, thank you. And, um, you said you're just back from holidays, so, uh, thank you for giving up your time.
[00:31:06] Speaker 2: You're grand. You're grand. And- No worries at all. Thank you for having me.
[00:31:10] It's been, it's been lovely to chat about this.
[00:31:12] Speaker: Yeah. It's been, it's been really, really interesting. Mm-hmm. So thank you so much. And, um- Uh-huh ... we will see each other very shortly, and hopefully we'll do another podcast soon. Thank you. Sounds
[00:31:20] Speaker 2: good stuff. I look forward to it. Thanks very much, Katrina. Thank- Thanks for having me.
[00:31:23] Thank you.
[00:31:23] Speaker: Thanks so much, Val.
[00:00:00] Speaker: Welcome to the next episode of the Humanizing IT podcast, We Need to Talk About IT. And yes, we do. On that note, my next guest is someone that I admire. Uh, I met this guest probably about six months ago at a conference in the United Kingdom, at an international Woman's Day, actually, for IT. Uh, this guest was the emcee.
[00:00:22] I was super, super impressed with her ability to engage the room, and also her ability to really, uh, explain and help what's going on in the IT industry today. Our guest works for one of the largest corporations in the UK, and I'm proud to introduce Valerie Wilson.
[00:00:40] Speaker 2: Thanks very much. I'm delighted to be here.
[00:00:42] Thank you very much for asking me, Katrina. Mm-hmm.
[00:00:44] Speaker: It's, as I've said, a absolute pleasure, and you are the, as, the first woman we have had.
[00:00:49] Speaker 2: Very... I feel very honored- ... on behalf of all women. Hopefully I'll do it justice.
[00:00:54] Speaker: Yeah, but-
[00:00:55] Speaker 2: Hopefully, yeah ...
[00:00:55] Speaker: uh, it's interesting, and, and as I said, we, we like to take a bit informal, these podcasts.
[00:01:00] I did say to you, Val, that I have been, um, sort of pulled up to say, "Hey, Kat, how come you and Wes only have men?" Mm. And I do ask women, and, and most of them don't respond, or, and I'm not sure why that is. But when I ask a man, they jump at it. And so yeah, it's, it's, it's, you know, it's an interesting-
[00:01:21] Speaker 2: I don't know why that w- I don't know why that is.
[00:01:21] Mm. I, like I... I've done a few podcasts. Not loads, but I always, I al- I, I might be wrong, but I, I love, I love a chat, you know? I love a good chat, you know?
[00:01:30] Speaker: Yeah, no.
[00:01:31] Speaker 2: Work and what's happening in the industry- Absolutely. Yeah ... the chance to, to share with what's going on. Mm, mm. So suppose everyone's different, but yeah, I can, I can imagine some people, but maybe more so females, are a little bit reluctant- Mm
[00:01:43] do you know? 'Cause it can be a bit daunting, let's be honest- Yeah ... you know, that way. Yeah, mm. You know, maybe you're going on and you're effectively just talking live about whatever it is that, you know, the person wants to talk about. So I can sorta get it. Yeah. But yeah.
[00:01:54] Speaker: No,
[00:01:55] Speaker 2: yeah. Like I said, it's down to the, the confidence thing,
[00:01:57] Speaker: you
[00:01:57] Speaker 2: know?
[00:01:57] It...
[00:01:57] Speaker: I, I agree. Yeah. Mm-hmm. So thank you for, for being the first. Yeah. Now Val, your role is Manage Services and AI Ops Program Director for British Telecom. Have I got that right? Mm-hmm. Fantastic.
[00:02:11] Speaker 2: Do you know what? I'm only, I'm o- I'm not even sure myself of my job title, but if that's what it says on LinkedIn-
[00:02:16] it must be true, isn't that right?
[00:02:17] Speaker: That's, that's what it says on LinkedIn. And- Yeah ... uh, if you could tell the audience a bit about you. Uh, uh, I said that, you know, the things you do personally are... You're, you're such- Yeah ... a busy lady, and professionally. Yeah. Just, uh, tell us a little bit about you.
[00:02:31] Speaker 2: Yeah. So, so I'm Valerie Wilson, for those of you who don't know me.
[00:02:34] Um, I, I live in Northern Ireland, in Belfast, so you can probably tell by the accent. Um, I think on a, on a personal level, first of all, I am married to Mark. I've got three daughters, Hannah, Jessica, and Amy. Um, so they're 16, 14, and 11. So some people say good ages and then some people say bad ages, but yeah, I, I quite like that age.
[00:02:52] There's less having to be there holding their hands, but yet more importantly emotionally probably, um, for girls at that age. So that's the kids. I, as you say, I, I pack a lot in to, um, I, I do quite a bit of sport and fitness. Um, this season again I've started to play football again, as well as coach football.
[00:03:11] Um, so I've always played football over my, my lifetime, and I sort of went into retirement to coach, and then the club I'm at started up a senior ladies and they're like, "Val, we need you to come back and play." So I've dragged back into actually playing football and in mid-40s, it's not probably ideal, but there you go.
[00:03:26] We're still keeping up with the kids, shall we say. But yeah, so I play a lot of football. Um, I also run a youth club, um, for our local church, and um, I sit on the board of governors, and I also sit on the board for ITSMF UK as well, and chair the women's network. So I suppose from a personal point of view, I do, as you say, I, I'm one of these people that if there's something needs done, I'll be like, "Yeah, I'll do it."
[00:03:49] Or if there's a committee needs done, I'll end up running it and I need to stop because my, my days are becoming jam-packed. And then as you say there, I have, um, a really exciting job at work at the minute as well. So I've worked for BT for 25 years, my whole career, um, in BT. Um, so I've done a lot of different jobs over that time.
[00:04:07] Probably for the last 10 years, focused more on, um, IT service management, managed services. Mainly across my career has been leadership roles, but this most recent one, which is only since January this year, is all about implementing AI into IT service management and how can we do things better, quicker, faster.
[00:04:25] So a lot of buzzwords in there, which means I get a lot of cold calls now than I ever have done in my life, do you know, from, from, from LinkedIn, and I wonder who got my number, um, all the time. But, uh, yeah, it's a r- I'm really fortunate to have this role, um, I have to be honest, and it is exciting, but it is, it is tricky, but it's good.
[00:04:42] Mm. It's good. It's all good.
[00:04:43] Speaker: Mm, 'cause yeah, AI, AI, AI, that is-
[00:04:47] Speaker 2: I know.
[00:04:48] Speaker: Tell me about it ... everything we hear. And, and it's interesting your role, but if you could explain, uh, 'cause I think it's a, a new, I, I just did an a, a podcast with Akshay, and, uh, we were talking about how AI is impacting roles in IT service management.
[00:05:05] So what is AIOps?
[00:05:09] Speaker 2: Yeah. So really if, if you break it down and ke- try and keep it really simple with AIOps, it's using big data. And when I say big data, it's big, right? It's, it's more data than you can really use in terms of every day. So what you're using all types of feed from a telemetric point of view, and then you're running machine learning over the top of that effectively to do things quicker, better, faster.
[00:05:31] Yeah. So it's that combination of big data, machine learning, and then your automation and your AI over the top of that. So the way I, I try and keep it simple and go, you think of all of your feeds that you have today that gives you information, just to keep it really simple. You bring them all into one area, and then you run effectively your machine learning over the top of that, and then you use the different types of AI to really try and pull to the top what are the best processes, the best, most repeatable tasks based on the history that we've seen, and indeed what we think is gonna happen in the future, and how can we run that quicker and better.
[00:06:04] That's probably, in my opinion, the easiest way to describe it. But yeah. In the operations, it's in that operational area.
[00:06:11] Speaker: But how would that work? 'Cause we, we did talk about this earlier, that you're focusing on AI for IT service management versus-
[00:06:19] Speaker 2: Yeah ...
[00:06:19] Speaker: AI for BT business, which I think is fascinating. Yes.
[00:06:23] Yeah. W- In, in regards to IT service management, what big data do you get? I mean, is it incidents or, um-
[00:06:29] Speaker 2: Yeah. It's, it's across a lot of d- of different things. So as you say, I'm very, at this stage, I'm very specific at managed services. And when we say managed services, what it means, and this can mean different things for different people- Yeah
[00:06:39] but in, in, in my context, it's effectively where do we proactively monitor our customer's infrastructure, right? So BT, it's not really a desk, you know, it's not really let's look at, at the internal IT. We run our platform across thousands of, you know, external customers, um, and it's across any technology.
[00:06:58] So it could be network, could be, um, firewalls, it could be data, it could be voice, it could be, um, server storage, it could be anything at all. So our big data, back to your question, effectively comes from all of our different tools that we use to monitor, um, the performance of the customer's technology. So it, it comes from probably a range of different things.
[00:07:17] And equally because we're BT, we obviously run the core network as well. Mm. So we also take feeds from the core network itself. So we've got feeds from the core network and from the customer's network- Right ... whether across all their technologies. So really then you're getting this huge lake of data that, that we're able to then interrogate using the AI, the machine learning, to try and do things better and quicker and faster.
[00:07:41] So in simple terms- I, I always, when we talk about a managed services, I, I describe it as Big Brother, right? So Big Brother is watching you, i.e., we are monitoring. Uh, so managed service and monitor, they're the two things. So M and M, managed service, monitor, and that's ... And if it's not monitored, then that's just a service, not a managed service.
[00:07:59] Speaker: Right.
[00:07:59] Speaker 2: And then you don't have as much rich data-
[00:08:02] Speaker: Mm ...
[00:08:02] Speaker 2: effectively to be able to run your machine learning, your AI over the top of. But because we run a managed services, and because we've got lots of different tools in our stack, we actually are in a real, I think it's a real advantage to have that information and then use that effectively to do things better and quicker and faster.
[00:08:18] Speaker: So when, when you say that, and again, I'll, I'll go back to service management, and this might be a stupid question- Mm ... but typically-
[00:08:27] Speaker 2: There's no such things. Might, might be a stupid answer, to be honest with you, Katrina.
[00:08:30] Speaker: Who knows? Who knows? And then this is a cynic coming out of me, Val, but typically when an IT organization embark on a, a new -- If we take AI out for a moment, embark on a new tool or a new transformation- Yeah
[00:08:41] we start with the usual suspects: incident, problem, change, and service request. Yes. Is that what you start with?
[00:08:48] Speaker 2: Yes, absolutely.
[00:08:50] Speaker: Oh, do
[00:08:50] Speaker 2: you? Absolutely. It i- Yeah, it is. I know it's not a stupid question at all. I, I, I tell you why.
[00:08:55] Speaker: I was hoping for a different answer.
[00:08:57] Speaker 2: No, I, I'm sorry. It's not a different answer.
[00:09:00] Maybe that's the wrong answer, but it's not. I, I'll tell you why. Um, whenever we've looked at our business case around doing it, you know, we have to look at where the bulk of the effort is, right?
[00:09:10] Speaker: Okay.
[00:09:11] Speaker 2: So whenever we looked at where the bulk of the effort is being spent in our IT service management world, as you say, it's incidents, it's service requests, and it's change, right?
[00:09:21] Now, ultimately, let's say whenever you, you've finished that bit and you wrap a bow on it and it's all nicely done, which dear knows when that will be, of course you would start to look at those wider ones. Do you know, you would start to look at capacity management, service continuity management, asset management, right?
[00:09:35] You would look at all those ones, but the first ones, and it may just be the way that we're set up and our effort, but the first ones were the most effort. And also from a customer perspective, you probably want to improve first is your incidents, your SSRs, your change. Do, do you know? There you go. It's not, not rocket science.
[00:09:52] Oh,
[00:09:52] Speaker: dear, oh, dear. And here I was thinking that AI's gonna change how we do things. But why don't you, um, start with problem management?
[00:10:02] Speaker 2: Yeah. So, it, it's, and there is obviously inferred benefits to problem management, right, when you look at incident. But the reality is, a- and, and again, I think if you had an endless pit of money-
[00:10:13] Speaker: Yeah
[00:10:13] Speaker 2: Katrina, right? You have to be able to prove your return on investment quite quickly. And I don't know if it's just a BT thing, but our rules are pretty stringent about CapEx and investment. So you have to be able to demonstrate the payback as well as, of course, do you know the benefits for the customer- Yeah
[00:10:31] do you know, and everything else. And therefore, the effort is involved. So our problem management team is relatively small in comparison to our incident management team, and that's probably as a result. Now, don't get me wrong, there are definitely some what we call quick wins or out of the box areas that would impact problem.
[00:10:48] So we're not saying we're not touching problem, but it's probably just not a radical, do, do you know what I mean? Let's start with problem first and not look at anything else. I think that would definitely be wrong. But to give you an example, um, in incident management, one of the, one of the use cases we're looking at is around auto of closure codes.
[00:11:02] So how do you get the closure codes right every time? Now, ultimately, if you do, if you get that right through, um, machine learning, right, and as we know, people at the minute when they do their closure codes, you know, you sometimes you get a bit trigger-happy. Right. And you go with whatever's the easiest thing, everybody not quite sure, and that obviously impacts problem management 'cause they're working- Sure
[00:11:19] off data-
[00:11:20] Speaker: Mm.
[00:11:20] Speaker 2: Right ... that's not real. Mm-hmm. Whereas whenever you use AI for incident to get the closure code right, ultimately helps problem management. Similarly with the ticket summarization to know how do you summarize things quicker, that helps problem m- whenever they're looking at their RFOs to know when their PIRs that they're trying to do post investigation around that.
[00:11:39] So, so it's not that we're not doing problem management- But there's elements of incident management- Mm ... that will make problem management a lot slicker around that. Um, and we do have some further use cases, and this is probably another thing around the data. So problem management relies on the data, and with any AI program, your first year really you need to get your data right.
[00:11:58] You know, your structure right, your governance right, your correlation right. Um, and after that you can start to see benefits flowing through into other areas where maybe you can interrogate the data be- better. You know, there's a great tool that we're putting in, which is around, um, how do you get good business insights from the data?
[00:12:14] A bit like how do you use Copilot over all that big data that we're talking about. Mm. Now, that will massively help problem management, but you can't do that-
[00:12:21] Speaker: Yeah, unless you have the data ...
[00:12:22] Speaker 2: your first probably year, year and a half- Mm ... until you get all of that in place- Mm ... and everything right. Mm-hmm. So there's probably gonna be more implied benefits down the line that we maybe aren't even know ourselves yet until we get that- Mm
[00:12:32] that part done. But yeah, that's probably ... I would say it indirectly impacts for a benef- better problem management, but I don't think, I don't think you would ever just go, "Let's look at problem management first," and do it. Do you know? I think you would probably do it the way we're, we're looking to do it.
[00:12:46] Speaker: It, it's a, an interesting view that I'm kind of embarrassed that I really haven't thought of this before. In some ways, um, you need incidents to do better problem management.
[00:12:57] Speaker 2: Well, that's some very interesting incident. There's no, there's no other way. That's a ... Well, let's ... That would be utopia, wouldn't it?
[00:13:03] Or, you know, it would be absolute utopia if there's no incidents.
[00:13:06] Speaker: Well, that's right. I don't know if that ever happens. So next time there's a major incident, um, everyone can thank you, Val, because y- you, you're just saying you need the data.
[00:13:15] Speaker 2: That's exactly right. Exactly.
[00:13:17] Speaker: And, and again, within your, um, experience and obviously the research you're doing with your role, um- What's the hardest part about AI, implementing AI for- Right
[00:13:30] IT service management? Do you implement it? Is that the word, or what, what is the word you say?
[00:13:33] Speaker 2: Yeah. Yeah. Well, yeah, you could say that. Yeah. I mean, 'cause a right or wrong word to say- Yeah ... but probably implementation's the, the right word. Oh, man, th- there's, there's quite a lot, and, uh, t- I, I think at, at different stages of your program, depending on the maturity of the program, there'll be different challenges.
[00:13:47] Um, I think initially the biggest challenge was getting a business case, you know, that worked, you know- Oh ... in terms of getting that payback that we talked about there. It gets death have been difficult. Hmm. Um, I would say as we were progressing through, if you like, this data governance stage, that's another big problem, do you know, is making sure your CMDB, your inventory is good and, and with all of these programs, and you'll hear it over and over again, and now that I'm in it, it is so true, that first year of, of any AIOps program, there's very little, I'm gonna say wow moments, because the data has to be right before you can get, get to the good stuff.
[00:14:24] And a lot of people, 'cause to your earlier point, AI is the big buzzword, everyone wants to know, like I spend most of my time updating, you know, s- more senior people than me about how the program's doing, and there is a impatience, let's say-
[00:14:36] Speaker: Hmm ...
[00:14:36] Speaker 2: do you know, you know, from people who maybe aren't so close to it or don't need to be close to it, whereby they've gone, th- they've given you the check, do you know, and they wanna see the results there and then.
[00:14:47] But that first year, year and a half, there's so much work and effort required to do on your data structure, your CMDB, and making sure that is right. Because if you don't have that correlation right, as I say, all that good, do you know, to be in AI or the, the more sexy technical stuff that people expect and you hear about AI doing, y- you can't get to that.
[00:15:07] So I think there's an element of that first year being tricky to get the data right and the expectation. And the final thing I would say is around adoption, do you know, um, value realization. Um, and again, it's maybe it's the stage of the program we're in now- Hmm ... but whenever you're implementing, you know, some of the, let's call them the quicker wins, you know, you are relying on people, you know, you know, I'll need to hit this button to make sure this summarizes the ticket, or I need to hit this button to do a, you know, to, to craft the email that's gonna go out to keep the customer informed.
[00:15:39] There is an adoption piece there with people, and I think that fear factor kicks in, d- you know, um, around that in terms of I've always done it this way, and how do, how do we do that? And it's not like- It's not like a, you need a training course to do it 'cause it's pretty straightforward. Do you know, you hit this button.
[00:15:55] You know, it's not like you're taking them all- Yeah ... you know, away and you're going, "Right, let's sit in the room and let's explain this," because it's not difficult per se. It's more of a psyche thing, do you know? And, and, and I get it, and even if you think of, like, some of the stuff we're doing is around, you know, how do we get our auto diagnostics, do you know, everything there if you like on a page for people to see.
[00:16:13] There you go, boom. But people do like to run it themselves. People like just to, "I'll just do a wee check here and I'll run it just to make sure in the background," do you know? Well, act- it's there. You don't need to, but, you know, that, it- it's that, it's getting people out of that way of thinking, and I think that will take time.
[00:16:29] Do you know, that's not something that happens overnight whenever you go live. It's something that ramps up, do you know, over time. Uh, and, and that's a tricky, that's a tricky one as well. And- They would be my three.
[00:16:41] Speaker: It's, it's interesting. Um, I'll, I'll... There's lots of points in there, but if I go back to that, um, dirty word called CMDB.
[00:16:50] Speaker 2: Yeah.
[00:16:51] Speaker: Is that important?
[00:16:53] Speaker 2: Oh my word. Yeah. Like, honestly, you, you, you can't underestimate the importance of that.
[00:16:58] Speaker: Isn't that, isn't that interesting?
[00:16:59] Speaker 2: And it depends how ... I wanna say, and it de- like, I don't know how, what other companies are like with their data, but anyone I've ever spoken to says the same thing.
[00:17:06] That no one's data's brilliant, right? Let's be honest. Yeah. And a lot of that, a lot of it is historic. And certainly from a BT perspective, it's an historic problem. It's usually because of a, a full lead-to-cash journey, you know, problem. It's not necessarily something, you know, that you've got wrong, uh, you know, once something goes live.
[00:17:23] It's right from when something is ordered, delivered, that whole process. Yep. You almost need another whole transformation program just getting that smooth, which we have as well, you know, in BT. But it's, it's not an easy thing to fix that, and we've tried to, um, look at, well, how bad do we think it is at the minute?
[00:17:40] And, and it wasn't as bad maybe as what we thought it was. Mm. But you need to work out quite quickly what are the minimum data sets you require to make all your AI run, and therefore how mu- how much are we gapped on? And therefore, how do we get those, how do we get those- Mm ... filled, do you know, that way- Mm
[00:17:56] to make sure it's in the right way. But we've been working very closely with, um, other people who've done this before. Yeah. I think if you were to go do it right on your own without real expertise there- Yeah ... you'd struggle. So to be fair, ServiceNow have given us, um, a lot of their more, I'm gonna say, expert resources- Right
[00:18:13] to know where they're really helping, helping us to drive the value in that, where they've done it before for other people. The same with Accenture. You know, they've done it before for other people. So they're trying to take ... Albeit our ecosystem and our infrastructure is very different, but in terms of the principles, they, they, they've learnt the lessons the hard way.
[00:18:31] So we're hoping, you know, that that's gonna help, and we're not gonna be here in a year's time and go, "Flip, our, our data's still not right, and our CMDB's still not right." Which I think you could find yourself quite easy in that position if you don't really go at that, I suppose, with the reverence that's needed to go, "This is the most important thing that you need to get right to get the benefits down the line, and make sure it's working down the line."
[00:18:54] You know? It is. And I was chatting to-
[00:18:55] Speaker: Mm ...
[00:18:55] Speaker 2: another lady, Siobhan, um, Hobbs, the other week- Yes ... today, and she was doing something. You know Siobhan?
[00:19:00] Speaker: Yes.
[00:19:00] Speaker 2: And it really encouraged me actually, 'cause she says as well- Year one, you're just scraping the surface. You know, if anyone thinks there's any benefits in year one- Yeah
[00:19:07] they're, they're really, really deluded. Yeah. And I think I'm so glad to hear you say that , you know? Yeah. So glad, because you know, it is, it's year one is a
[00:19:16] Speaker: difficult
[00:19:16] Speaker 2: year- Yeah ... you know, to get through on that, on that front. And, um, I, I think to over-perform in year one in the AIOps program is probably the hardest thing you'll ever have-
[00:19:26] Speaker: Mm
[00:19:26] Speaker 2: ever to try and do. Yeah.
[00:19:28] Speaker: But it, it's kinda like saying, um, frameworks like ITIL, they've always had this governance in play for service management, just, just CMDB and your hygiene of tickets. Yeah. It's kinda like saying, "I told you so." Like you, you need that
[00:19:44] Speaker 2: to- Yeah. That's so true It's wise. It's so true. And people, you always hear people saying things like, "Oh, if you're building a house, you know, you need the foundations, and the foundations is CMDB." The amount of analogies you've heard about CMDB- Yeah, yeah ... and houses, and foundations. Yeah, yeah. Blah, blah, blah.
[00:19:54] Yeah. Do you know? And I'm trying, I'm trying to get the furniture in and haven't even built the walls. Do you know, it's, it's thing, it's things like that- Yeah ... that you hear all the time on different areas, and it's so true.
[00:20:03] Speaker: Yeah.
[00:20:04] Speaker 2: Um, it's so true, but I think AIOps, if you're doing it in your business, this will force that.
[00:20:09] It will absolutely force- Yeah ... your hand to do that. Yeah. And I think that's not, not a bad thing. That
[00:20:14] Speaker: thing. No, a- absolutely. That's a good thing. Because, yeah, I mean, that's why we have these, this governance in service management- Mm-hmm ... is to have a, a smooth and incident-free, uh, or incident reduction environment.
[00:20:24] And Val- Yeah ... w- in relation to, um, I guess day-to-day processes and how we know them- Yeah ... today, we do knowledge management, problem management, and we log a ticket. Yeah. We might functionally escalate it, we give it to a resolver group. Yeah. Will that fundamentally change, do you think?
[00:20:43] Speaker 2: Will it fundamentally change?
[00:20:45] I think what will fundamentally change is the effort required to do that, but I don't think ... Is it gonna fundamentally change? I, I don't think it will, right? Look, uh, potentially, as I say, it'll be a lot quicker than what it's done now, today, but ultimately, ultimately those steps are still the right steps.
[00:21:07] You might shortcut a few, do you know, that way? So for example, I'm just trying to work through an example in my head. Can
[00:21:11] Speaker: I, can I give you one?
[00:21:12] Speaker 2: Yeah, go for it. You go for it. Yeah.
[00:21:14] Speaker: So it, it's interesting you say that, and sorry to interrupt, but part of what we do at Humanising IT is to challenge- Mm-hmm ... what we do today.
[00:21:22] Mm-hmm. Yep. Because I think, um, you know, we talk about something called IT service management confirmation bias, and-
[00:21:30] Speaker 2: Yeah ...
[00:21:30] Speaker: we have a, the quiet architecture in IT service management, meaning that we're following best practice. And, and I always give the example- Yeah ... contact the user three times before you, um, close the ticket.
[00:21:42] Yeah. Now, to me, you know, we've never challenged that mostly in, in IT service management. Yeah. We think we're doing the right thing. Um, another example is close the ticket if a workaround has been provided. Um- Yeah ... so that's why I'm asking, and we sort of see within Humanising IT, AI or any improvement really, gives you the opportunity to say, "Hey, stop.
[00:22:04] Is this actually the most efficient-" Yeah ... efficient way to, to do it?" Um, just 'cause we've always done it- Yeah ... doesn't mean it's, it's the right way to do it. And, you know, we quite often say, "Did anybody ask the user if they wanted to be contacted three times?" And the answer's invariably no. So true. So that's what I mean.
[00:22:21] Yeah. Like, you know, are we using- Yeah ... are you using the time to sort of- Humanize processes, um, or just fundamentally, you know-
[00:22:29] Speaker 2: Yeah. It's a really good, good point. So every time we look at a use case, um, uh, in terms of what we're gonna do, we walk through the process, and then we think about, well, what, what process do we want it to look like, right?
[00:22:40] Yeah. And then how do we use AI to improve it? So invariably, you do look at that, but, but what you'll probably find is people are so used to that way of working that it's quite hard to, to come out of that box. Yeah. Do, do, do you know? And what I would say with some of the, the big ITIL processes that we all know and love effectively, they are high level, right?
[00:22:58] They are, as we know, high level, and typically what it gets into is local work instructions, which is if you like- Mm ... the detail, detail below it. Yeah. So an incident, whatever, you've got your five key steps, whatever it is. You know, but underneath that, in terms of that local work, and I think those broadly speaking, the five stop steps at the top- Yeah
[00:23:14] do you know, are, are right. But whenever you get into actually what happens on the ground- Yeah ... how do people use those steps to navigate through it? Mm. That's the bit I think that needs probably to change- Yes. Yeah, yeah ... do you know, more so than your broad, high level- Yeah ... bit at the top. Like, it would be lovely to get to a stage where you never need a CAB, for example, in change.
[00:23:32] Mm. Sh- show that way. Mm. Because everything has been de-risked down to, rather than a complex or normal change- Yeah, yeah ... do you know, to it, it's a standard change. Change, yeah. Because of, because of the way AI will analyze a change. Mm. Do you know, it will be able to, with, with confidence, say what the impact is.
[00:23:50] Mm. With confidence, say, actually, there's a, there's a, a lack of resiliency over here in advance, so we need to make sure that that's looked at. Do you know, we've already done the rollback plan, we've done this, we've done that and the other. You know, it should be how do we get more changes to be standard, do you know- Mm
[00:24:03] therefore not needing- Mm ... to go through the CAB, do you know, to be able to do that. Um, a- and how do we then work that through? So- I think in Verb, and there's definitely gonna be different skills for people required to do all of these things, and I think what you may find is where elements are going to different resolver groups today, they may not have to go there in the future.
[00:24:21] Yeah. Because AI will go, "This is next step. This is what I think we should do next based on the last thousand tickets that look like this- Yeah ... for this customer." Like, you know, "No, no surprise, guys," do you know what I mean? "We're gonna need to do it like this. Therefore, do you want me to do it like this?"
[00:24:37] And the person there will look at it and go, "Yes, I do." Get on and, and do it. Yeah. Whether that be a change to the VLAN or, you know, changing a config or reboot, whatever it happens to be for, for that thing. Um, so you may find you don't have to use those user groups as much, or indeed vice versa. You may find it goes directly to those user groups- Mm
[00:24:55] because it knows based on the history that every time someone at the frontline gets it, that's where it goes to. Yeah. So it cuts out, if you like, the middleman. You know, the middleman could be anywhere in these situations, but ultimately it will, it will get the same outcome, right? Mm. But potentially a lot less steps.
[00:25:12] Yeah. Do, do you know? Certainly a lot less steps from a human touch point of view, I think is probably the way I can see that going down the line. A lot more about people understanding the guardrails and, if you like, checking the work that AI, or checking the next steps that AI is saying. Do you know? So you're always, I think, gonna need a human in the loop somewhere there on, on, particularly with the customers we're working with, which are very much your cr- critical national infrastructure customers.
[00:25:37] Um, so I think that's an important factor for sure.
[00:25:40] Speaker: And do you think the days, it's, um, probably a stupid analogy but I'll go with it. Um, I, I spoke to someone from America recently, and we actually had a meeting, um, and he was in his drive- driverless car. And-
[00:25:55] Speaker 2: Okay ...
[00:25:55] Speaker: you know? And you know, I mean- In the
[00:25:57] Speaker 2: meeting, right.
[00:25:59] Speaker: And-
[00:25:59] Speaker 2: That'd be a bit controlled ...
[00:26:01] Speaker: you know, and, and you know, we, we're fortunate enough to travel, um, the globe and we speak at- Yeah ... conferences and in, in Vegas and, you know, other countries in the world. Yeah. You know, they tr- have true driverless cars. Yeah. And, and to me, I, I think that technology is amazing, you know?
[00:26:14] Yeah. Um, and is service management gonna become like a driverless car, where we don't need people to steer and direct? Uh-
[00:26:25] Speaker 2: And I, I, oh, I, I don't think so. Okay. I honestly don't think so. Interesting. I really don't. I don't. Um, not ... A- and it may just be that it's just too far away at this stage to, to, for it to be a reality.
[00:26:39] I think there's absolutely elements right today where we have humans doing things that we won't need them for, but as a collective and as a whole, when you're running a managed services, the humans are so important, right? And, and that's because customers are so important ultimately at the end of the day, and customers need relationships.
[00:26:58] They need to build trust, and they need to have people like they know are doing that. And like we have, um, we haven't done, I would say, a m- massive launch externally. That's gonna be coming from a BT perspective, but we have spoken to a number of customers to know about what we're doing, how we're doing it, and there's definitely a nervousness, a nervousness and excitement around that.
[00:27:18] But what they will always ask is, "Hold on a minute. How are you gonna prove that, you know, AI knows what it's doing? How are you gonna prove that with confidence you need that?" And what reassures people at this stage, in a way, is that human in the li- human in the loop model. You know, we've got a human there checking, making sure, and I think the governance on the AI is so important.
[00:27:39] You would have to be proven, right, this is 99.99%- Right. Yeah ... accurate at all times, do you know, that way, before you'd remove the human, and there will absolutely be circumstances where you would never, even if it's 100% every time, you would never- Right, the human ... remove that human in the loop because of the types of services, do you know- Mm
[00:27:57] that we are looking after from an AT perspective. Mm. The, the risk and the impact of something going wrong, and like, it's funny because over the years, like all the major incidents, do you know, like, uh, from, uh, Mem- oh, Mem- Memory Time, the majority of times a, a major incident happens because of human error Is mad.
[00:28:14] But yeah, we're talking about AI, and everyone's more freaking- Yeah ... about AI error, and I recognize humans have humans have error as well. Very much so. As a manager, you can see that in so many instances, so- Yeah ... it's quite ironic, do you know, the way, but I think even though they're probably, their error rates is probably same if not worse- Yeah.
[00:28:31] you know, people feel more comforted. Mm. But I think a- as much as anything, it's about how do you interpret what all of this data is telling us.
[00:28:40] Speaker: Mm.
[00:28:40] Speaker 2: And that interpretation, right, come, should be coming from your technical experts in managed services. Yeah. Mm. And playing that back to their counterparts with our customer to have those conversations to make, ultimately make better decisions for their future.
[00:28:54] Mm. And that's, that's what it is. Yeah. You're always gonna need humans there in various forms. A- as I say, I don't think you're gonna need all of the different roles that we have today. Mm. And I think it could be a whole different way of working in it, but during, I don't think, I don't think you'll ever get, "Let's do a managed services as a driverless car."
[00:29:11] Do you know, unless it's a very, very standard service- Yeah ... you're providing. Yeah. Do you know, more the SMD, you know, your, your very small medium business. You're not- Yeah ... getting into your bigger customers. Yeah. Do you know? Like, there's definitely an option for that for the low-end market, you know, where- Mm
[00:29:25] where, you know, a different price point, but when you get into your more complex managed services, I don't think that's gonna ever be the case, to be honest- No ... with it.
[00:29:34] Speaker: And, and to be honest, that's really great to hear, you know, someone like yourself being, actually doing- Yeah ... AI in, in ops, to hear that, that we still always need a human in the loop.
[00:29:44] Um- Yeah ... and it's not always, and it, and you know, good news too, it's not that roles, well, roles will go, but skills will change, and, and that's- Yeah. Absolutely ... again, an opportunity. So yeah. Mm-hmm. Uh, and, and just before we go, I did wanna mention that, um, you know, another reason I was so excited that you s- accepted to appear on the podcast, and I mentioned that we go to Vegas quite often, and the last conference we did was called Humanizing AI.
[00:30:08] Yeah. And, uh, we actually were at full capacity. Now we do training on, on this, and we normally have 20 people in a class, and unfortunately, might be my fault, um, the conference organizers didn't ask how, what's the capacity you can have. And anyway, we had 100. in a class and the demand, and they actually had to cut it off 'cause they didn't have the room but, and we had an...
[00:30:32] I won't mention what the other classes were, but they, they had the 10 and 20 in the class, and we were packed. Yeah. And it's just so interesting that this whole AI, you know, industry, the, the interest that we have, and I think that-
[00:30:45] Speaker 2: Yeah ...
[00:30:45] Speaker: there's not enough information for, for people, so that's why it was, s- really great for you to come on this podcast and actually- Mm-hmm
[00:30:52] what, what is it, and what's it like day to day? And, and for that, I'm, I, I appreciate, and I'm sure our, our listeners do also. Um, and yeah, so on that note, thank you. And, um, you said you're just back from holidays, so, uh, thank you for giving up your time.
[00:31:06] Speaker 2: You're grand. You're grand. And- No worries at all. Thank you for having me.
[00:31:10] It's been, it's been lovely to chat about this.
[00:31:12] Speaker: Yeah. It's been, it's been really, really interesting. Mm-hmm. So thank you so much. And, um- Uh-huh ... we will see each other very shortly, and hopefully we'll do another podcast soon. Thank you. Sounds
[00:31:20] Speaker 2: good stuff. I look forward to it. Thanks very much, Katrina. Thank- Thanks for having me.
[00:31:23] Thank you.
[00:31:23] Speaker: Thanks so much, Val.
