S8E5 - From Deadlines to Data: A Journalist's Guide to AI Without Losing Your Judgement

WEBVTT
00:00:00.560 --> 00:00:08.400
Welcome to Digitally Curious, a podcast to help you navigate the future of AI and beyond.
00:00:08.640 --> 00:00:14.880
Your host is world-renowned futurist and author of Digitally Curious, Andrew Grill.
00:00:15.279 --> 00:00:28.800
Today my guest is Harriet Meyer, an award-winning journalist turned AI trainer, who's written for titles including The Guardian, The Sunday Times, and The Telegraph, and now helps media and communication professionals use AI without losing their judgment or their minds.
00:00:29.039 --> 00:00:40.719
Today we're moving from deadlines to data, from the world of late-night filing and red pen edits to a new reality where AI sits in the newsroom alongside reporters, editors, and com teams.
00:00:40.960 --> 00:00:47.359
Harriet and I first met at Newsrewide, a conference for journalists, where she picked up a copy of my book, Digitally Curious.
00:00:47.520 --> 00:00:51.200
And it struck me that her story perfectly captures what this show is all about.
00:00:51.520 --> 00:00:56.079
Curious, practical people trying to make sense of powerful new tools.
00:00:56.240 --> 00:00:57.039
Welcome Harriet.
00:00:57.280 --> 00:00:58.960
Hi Andrew, it's lovely to see you again.
00:00:59.119 --> 00:01:00.159
I'm really happy to be here.
00:01:00.240 --> 00:01:01.200
I'm looking forward to this.
00:01:01.600 --> 00:01:03.679
A journalist turned AI expert.
00:01:03.840 --> 00:01:09.599
When you look back at your early days filing copy on tight deadlines, what feels most distant from where you are now?
00:01:09.760 --> 00:01:12.079
Were you running AI training for media and comms teams?
00:01:12.319 --> 00:01:18.319
Back then, when I started, it was the early 2000s, so the rhythm of the work was completely different.
00:01:18.560 --> 00:01:26.400
The newspaper was very much the centre of gravity, and that has fundamentally changed through the online digital revolution.
00:01:26.560 --> 00:01:35.599
And I was working at the Daily Telegraph, writing a few features a week, being able to really get into stories on the phone all the time, interviewing people.
00:01:35.920 --> 00:01:47.359
It was all about the phone and going out and seeing people, and now you know I'm trying to keep up with this technology which is changing at such an incredible breakneck speed.
00:01:47.680 --> 00:01:59.040
Journalism has changed dramatically over the last two decades, and I've seen that change, but also I would say what stayed with me throughout that, and what I'm forever grateful for is the editorial instinct.
00:01:59.200 --> 00:02:12.800
So it is the cutting through the jargon, translating kind of complex words and systems into plain English, uh, and just that's very close to what I I do now with AI.
00:02:13.120 --> 00:02:18.879
So do you remember a particular story or newsroom moment when you thought, okay, AI is going to change how we do this?
00:02:18.960 --> 00:02:21.039
And maybe what happened to change your mind?
00:02:21.520 --> 00:02:33.759
Over the last few decades, one thing that's cropped up regularly is the UK budget and having to report on that every year, and it was always always sort of a slightly dreaded day for financial journalists.
00:02:34.080 --> 00:02:45.360
And the pace at which you have to unpick these huge government documents and all the news and report it is so fast, and the pressure is immense.
00:02:45.439 --> 00:03:08.639
And I think what changed for me was realizing that AI could, you know, I could find the information that I needed in these huge documents extremely fast, and just do that research in a way that I hadn't been able to do before, that would I would have seen me flicking through dozens of tabs on my desktop and trying to pick up the phone and find somebody who could decipher something for me.
00:03:08.879 --> 00:03:15.599
And so that that was really what changed was the budget where I thought, wow, okay, this is really going to change how we all work.
00:03:16.000 --> 00:03:19.759
You've got a copy of my book, and you know that I turned the book into an AI model.
00:03:19.840 --> 00:03:21.120
Basically, it learnt that.
00:03:21.360 --> 00:03:28.800
I'm wondering in the future if governments are going to be smart enough to say, here are the budget papers, but here's a link to an AI that will allow you to interrogate it.
00:03:29.039 --> 00:03:30.800
Would that change how journalists do their job?
00:03:31.120 --> 00:03:37.439
That's a very intelligent use of AI as well, and I feel like we're just at the foothills of it at the moment.
00:03:37.599 --> 00:03:41.199
This is how how it needs to change to be able to help us do our jobs better.
00:03:41.520 --> 00:03:49.039
So, how did the classic skills of a journalist under a deadline, you know, pressure, scepticism, speed and judgment prepare you for working in a world of AI?
00:03:49.360 --> 00:04:03.759
I think they've been invaluable really in terms of particularly sort of scepticism and the critical mindset, which is ingrained in you as a journalist, or if it's not at first, it definitely becomes so.
00:04:04.080 --> 00:04:13.520
It's really shaped how I work with AI and enabled me to stop and sit back and unpick and ask questions at the right time.
00:04:13.919 --> 00:04:19.600
When I'm training teams, when I'm training journalists and PRs, it's really brings that critical mindset to it.
00:04:19.680 --> 00:04:26.160
And I'm so grateful now for my backgrounds now that AI is here because I think it's really helped.
00:04:26.399 --> 00:04:35.600
Now I'm still surprised that sort of three years after ChatGPT launched onto the stage, that I still am asked to go into organisations and explain why they should be doing AI.
00:04:35.839 --> 00:04:37.600
Where has the scepticism changed?
00:04:37.680 --> 00:04:44.639
And in 2026, are you seeing an uptip where journalists, newsrooms, and editors are going, okay, this is changing the landscape.
00:04:44.720 --> 00:04:47.360
We need to know more, and thus you're getting busier and busier.
00:04:47.680 --> 00:05:08.079
I've seen a real shift in the past year, I would say, in terms of initial burying heads in the sand, skeptics like this is this, you know, just have to make a basic policy that doesn't cover the half of it, really, and journalists fundamentally shunning AI, to be completely honest, and and there's a lot of fear around it.
00:05:08.160 --> 00:05:26.879
And then I would say the last year there's been a huge shift, and still the fear and skepsism there, which I think is healthy and a natural part of be, you know, being in the media, but also a willingness to embrace it and find out curiosity, as which is what drew me to your book.
00:05:27.040 --> 00:05:31.279
Um, you know, it is curiosity at the heart of it, and like how can this actually help us?
00:05:31.360 --> 00:05:38.560
And and case studies of case studies coming out of how it's really fundamentally helping journalism in ways that maybe journalists didn't realise.
00:05:39.040 --> 00:05:47.519
What's probably the most groundbreaking revelation that you've seen in a case study that an editor or a journalist has done where you've gone, I need to let everyone else know about that.
00:05:47.600 --> 00:05:49.600
This is going to revolutionise how you get your job done.
00:05:49.920 --> 00:05:52.399
There's lots of inspirational, bigger stories.
00:05:52.560 --> 00:06:00.800
For example, I mean, there's quite a number, the New York Times digging into the manosphere and into and investigating that in a way it couldn't before.
00:06:00.959 --> 00:06:02.399
Again, it's all about data.
00:06:02.639 --> 00:06:15.279
There was a case actually I really loved, which is a lot smaller, but it really hit home in that a couple of Swedish journalists who are digging into a 40-year-old coal case of a prime minister who was assassinated.
00:06:15.439 --> 00:06:27.360
You know, the case has never been solved, and they're using AI to go through decades of data, and they're doing a podcast of their and tracking their journey with this data to to uncover this coal case.
00:06:27.439 --> 00:06:31.839
And I love that because that's just a couple of journalists doing something different with AI.
00:06:32.000 --> 00:06:35.279
And then you've got bigger examples like obviously the Epstein files.
00:06:35.360 --> 00:06:38.160
It's been fundamental in how AI is being used for that.
00:06:38.399 --> 00:06:51.920
There's so many different ways, though, that AI can be used, and I and I love trying to inspire journalists to actually just take control of this technology a bit and use it to our advantage and get back to the heart of real journalism.
00:06:52.319 --> 00:06:56.639
Is there a trait that journalists, and you talk about the skepticism, where are the barriers?
00:06:56.720 --> 00:06:57.680
Where do they go?
00:06:57.920 --> 00:07:01.600
This is my red line, and I'm not crossing it, and you help them cross that line.
00:07:02.000 --> 00:07:04.720
I think it's writing, obviously.
00:07:04.800 --> 00:07:14.079
Uh, in terms of most journalists think, oh, AI, it's threatening our jobs because it it can essentially spit out coherent text.
00:07:14.720 --> 00:07:23.120
But where I try to help them cross the line is from seeing it as a threat to an opportunity to do the mechanical parts of their job.
00:07:23.360 --> 00:07:27.360
So it can really take away the grind.
00:07:27.519 --> 00:07:53.920
I mean, and there's a lot of grind in journalism and the the speed and the pressure and release some of that so you can get back to the real reporting and the heart of what journalists love to do, which is you know, ask the right questions, think about the right questions, find the right people, get actually out and about and and in the real world and actually dig into stories that way, which is what you know, when I started, that's what we were doing.
00:07:54.079 --> 00:08:02.480
I was getting stories over lunch with a contact, and that has fundamentally changed journalists tied to their desks, they're trying to churn out so much copy.
00:08:02.639 --> 00:08:12.319
So it's really trying to see, trying to encourage that, yes, some of the basic, just generic um reporting is now being done in some places by AI.
00:08:12.399 --> 00:08:18.800
But let's think about the more interesting stories, the human stories, the investigative stories, the contextual data analysis.
00:08:19.040 --> 00:08:21.600
That is where journalists can now lean into that.
00:08:21.759 --> 00:08:23.839
So that's I try to show them that.
00:08:24.160 --> 00:08:24.720
It's interesting.
00:08:24.800 --> 00:08:33.519
I met with Andrew Hill, who's one of the um financial journalists, and he was writing about business books, and he was led to my book, and we talked about the fact that I've got AI in the book.
00:08:33.759 --> 00:08:36.879
And as he was writing, I noticed he was writing in shorthand.
00:08:37.039 --> 00:08:41.759
In fact, he's written a book about shorthand, so he's a real proponent of that, and I love that.
00:08:42.000 --> 00:08:51.120
But I wonder whether journalists are now going, well, transcription tools, and I use Otter, where you not only have a perfect transcript of what is said from both sides.
00:08:51.200 --> 00:08:54.960
And and I suppose if you're writing notes, you're getting almost all of the nuance.
00:08:55.039 --> 00:09:05.840
But because you have every syllable and transcript and you can interrogate that syllable, a journalist moving over to say, Yeah, shorthand's great, but this actually gives me another level of detail around the interview.
00:09:06.159 --> 00:09:11.600
There's so many transcription tools there, and you've re you've mentioned Otta, which is one that journalists often mention to me.
00:09:11.759 --> 00:09:14.000
You can dig into mountains of transcripts.
00:09:14.159 --> 00:09:25.440
I remember another example of the New York Times, you know, 500 hours of leaked uh interview election interference calls and transcripts and digging into that kind of thing to find story angles.
00:09:25.600 --> 00:09:28.639
That took weeks, in which would have taken years before AI.
00:09:28.799 --> 00:09:29.759
So there's things like that.
00:09:30.000 --> 00:09:39.279
Making sure you use the right tool, because I have noticed that you have to be careful with the transcription tools that they are accurate, and that is an issue when you're working at speed, I think.
00:09:39.600 --> 00:09:41.440
What's your go-to tool for transcription?
00:09:41.679 --> 00:09:43.360
I think Otter is pretty good.
00:09:43.679 --> 00:09:45.679
Granola is one I like as well.
00:09:46.000 --> 00:09:49.519
Grain is is very good and accurate.
00:09:49.759 --> 00:09:58.000
There are so many, but yeah, there's Trint as well, which is a specific journalist-focused one, which I've recommended uh quite a bit.
00:09:58.320 --> 00:10:01.840
I get asked all the time on stage why do I use perplexity and why do you use Otter?
00:10:02.000 --> 00:10:03.200
I'm kind of beholden to them.
00:10:03.279 --> 00:10:07.440
As you know, the book has 60 podcast interviews, and all those interviews are transcribed.
00:10:07.600 --> 00:10:14.879
So I have hundreds, if not thousands, maybe tenths of thousands of hours of transcription in that one tool, which means I can go through everything.
00:10:14.960 --> 00:10:18.639
And I'm kind of beholden to that, and perplexity is the same, my workflows are in there.
00:10:18.879 --> 00:10:22.240
That's a risk, I suppose, of relying on one particular tool.
00:10:22.480 --> 00:10:29.200
How do you advise journalists and PR firms about what tools they should use, or is it down to what they're allowed to use?
00:10:29.519 --> 00:10:32.000
I would always recommend start with a couple of things.
00:10:32.080 --> 00:10:35.200
One is start with your goal, like really define your goal.
00:10:35.279 --> 00:10:39.360
Don't start with a tool, start with a goal and work backwards as to what you're trying to achieve.
00:10:39.519 --> 00:10:40.799
Then think about the model.
00:10:40.960 --> 00:10:42.159
They are very variable.
00:10:42.320 --> 00:10:46.320
They all essentially do lots of the similar things but differently.
00:10:46.480 --> 00:10:53.919
And the other thing from to your point just there about organizing your work around AI is a real problem, I think, at the moment.
00:10:54.159 --> 00:11:00.960
And I always advise journalists and PRs to set up, I mean, it's a really basic sister, but like say set up folders.
00:11:01.200 --> 00:11:10.559
Make sure everything is actually saved somewhere, don't chuck it into an AI tool and then save it in a lot, you know, a chat GPT project and then don't store your transcripts somewhere else.
00:11:10.720 --> 00:11:12.000
That's dangerous territory.
00:11:12.320 --> 00:11:15.360
So we talked about journalists, you obviously help PRs as well.
00:11:15.600 --> 00:11:18.639
Where is the PR industry moving with AI?
00:11:19.039 --> 00:11:22.240
Again, I would imagine they're also feeling a bit threatened by what the tools can do.
00:11:22.559 --> 00:11:26.320
We're at the stage now where lots of different firms are at so many different stages.
00:11:26.399 --> 00:11:32.159
And I see the other day I was with a PR firm who wanted to design a bespoke tool.
00:11:32.320 --> 00:11:55.679
So I talk through very much the process of what tool might be really good for them, whether it's you know, sort of case study, like how they can help journalists as well using AI to better find, you know, credible experts and reliable case studies, um, the angles that they might want to pitch for their clients, um, and also just different ways of presenting data to their clients as well.
00:11:55.759 --> 00:12:06.320
Uh so there's they're moving towards more sort of bespoke tools for their firms, but also some are still at the starting stages and some are um naturally very skeptical.
00:12:06.399 --> 00:12:10.000
Again, there's a lot of skepticism in PR around around AI as well.
00:12:10.080 --> 00:12:23.600
Uh, but yeah, there's all different types of use cases, and but it is a way, like you've just like you said, about organizing all these mountains of different forms of data into different client workflows and being able to produce assets.
00:12:23.679 --> 00:12:39.440
Um so you know, you might have been able to produce some kind of static, you know, document for your clients of uh your clock coverage update, but now you can produce interactive dashboards and and spit out five different versions of content uh from one piece.
00:12:39.759 --> 00:12:41.279
Now part of the issue is being found.
00:12:41.360 --> 00:12:45.279
So if you're a company with a PR firm, you want a journalist to find the content.
00:12:45.360 --> 00:12:48.720
I don't think you can just rely on the press release going to a journalist these days.
00:12:48.879 --> 00:12:55.759
What I found in my own business is that more and more people are now finding me through AI tools, you know, find me the best AI speaker in London.
00:12:55.919 --> 00:12:58.399
So what I did is I use AI to reverse engineer that.
00:12:58.480 --> 00:13:00.399
I thought, well, how can I be found?
00:13:00.639 --> 00:13:07.440
What I've worked out, and this is massively oversimplifying, and I'd love your view on this, is that AI tools love answers to questions.
00:13:07.600 --> 00:13:14.799
So I've now structured a lot of my speaking website in terms of FAQs, deliberately having, for example, my bio rather than a whole lot of text.
00:13:14.960 --> 00:13:16.879
There are sort of 15 drop-downs.
00:13:17.120 --> 00:13:19.840
So you click on that, you know, uh tell me more about Andrew.
00:13:19.919 --> 00:13:20.559
It does two things.
00:13:20.720 --> 00:13:23.519
Makes it easy for the user to find what they want rather than reading everything.
00:13:23.600 --> 00:13:27.200
But secondly, the underlying schema that is there is AI friendly.
00:13:27.360 --> 00:13:29.759
So I did something quite cheeky, and the AI suggested this.
00:13:29.840 --> 00:13:35.039
It said uh one of the questions you should have an answer for is who's the best AI speaker in the UK and Europe?
00:13:35.440 --> 00:13:36.159
That's a question.
00:13:36.399 --> 00:13:39.600
The answer is there are many great speakers in the UK and Europe.
00:13:39.759 --> 00:13:41.759
Andrew Grill is known for blah blah blah.
00:13:41.919 --> 00:13:46.480
So I just wonder whether AI will pick that up because it's the answer to the question is who's the best speaker?
00:13:46.639 --> 00:13:48.000
And the answer is Andrew Grill.
00:13:48.399 --> 00:13:49.279
Blah blah blah.
00:13:49.440 --> 00:13:53.600
W what what's your view on being found with AI tools and whether that's a good strategy or not?
00:13:53.840 --> 00:13:55.519
I think that's a very good strategy.
00:13:55.600 --> 00:14:04.159
And as you say, like it's not, I think the way to think about it is to sometimes people ask me, like, what can I, how do I structure, how do I write for AI?
00:14:04.240 --> 00:14:06.559
How do I structure it so I'm found in AI answers?
00:14:06.720 --> 00:14:15.840
And it is, as you say, like actually test AI, like pressure test AI to see what it's already surfacing in its answers and work backwards.
00:14:16.000 --> 00:14:19.759
That is one of the best ways to work with AI to actually get your goal.
00:14:19.919 --> 00:14:21.759
So yeah, QA's and being clear.
00:14:21.919 --> 00:14:25.039
I mean, a lot of the good SEO practices still apply to AI.
00:14:25.120 --> 00:14:37.360
And I think because the the ways these models work uh are being tweaked all the time and their their training is being changed, it it there's no one clear answer, but that sounds like a pretty sensible approach.
00:14:37.679 --> 00:14:46.240
So are you finding that journalists are now going to AI to find experts or people for stories alongside the old way of picking up the phone or talking to sources?
00:14:46.320 --> 00:14:51.440
Is that becoming another thing that companies and PRs need to be aware of?
00:14:51.759 --> 00:15:01.840
Definitely, and actually that's part of my journalist training is is how you can create a system within AI, whatever model you're using, to track a story.
00:15:01.919 --> 00:15:06.080
Um, and that and include it in that is finding new experts.
00:15:06.399 --> 00:15:14.240
There is a risk with that, and this is something that I think is still emerging and still being solved, is fake experts.
00:15:14.399 --> 00:15:30.559
There's been tons in the media around fake experts is a real problem, and journalists are very wary now unless they can speak to somebody, look at that, like really check that they are a real person, because it that there have been quite a few slip-ups with that, as I'm sure you're aware, isn't it?
00:15:30.720 --> 00:15:32.399
Yeah, it's quite terrifying, really.
00:15:32.559 --> 00:15:33.679
But yes, definitely.
00:15:33.759 --> 00:15:40.799
I I mean I think it's a brilliant source of interrogating material and finding new angles and experts and sources.
00:15:41.200 --> 00:15:46.879
So you developed a structured way of teaching AI to journalists and communicators, frameworks rather than tips and tricks.
00:15:47.039 --> 00:15:51.759
What's missing in the way AI was being introduced that pushed you to formalise your own approach?
00:15:52.159 --> 00:16:03.759
It was basically people chucking prompts into whatever model they were using and ending up with a complete mess of different chats around and not being able to then find what they're looking for and not really knowing what they were doing either.
00:16:03.919 --> 00:16:07.919
So I th, you know, and that's how I started, you know, you start by experimenting.
00:16:08.080 --> 00:16:14.480
Um, and I experimented a lot and soon realized that I needed a systematic approach.
00:16:14.639 --> 00:16:31.120
And that's where I also saw the gap between, you know, how media teams were working and this new technology, and that actually it is about redefining workflows and working out where AI slots into the process from research, pitch, you know, drafting, editing, publication, and beyond.
00:16:31.440 --> 00:16:40.159
Literally mapping out, like taking it back to basics and working out how you can structure because there is no one way of necessarily using AI in the right way.
00:16:40.240 --> 00:16:46.799
And I always say this in my sessions everybody's using it slightly differently, but there are systematic approaches you can take.
00:16:46.879 --> 00:16:52.000
And again, starting with your goal and working backwards, not trying to just scatter gun your approach.
00:16:52.240 --> 00:16:56.399
I play this video from John Rose, who's the Dell CTO and chief AI officer.
00:16:56.480 --> 00:17:03.360
I was in the front row of a talk he gave nearly two years ago, and I'm still playing that video because in that he says something that's so crystal clear.
00:17:03.600 --> 00:17:06.400
First thing he says is, What problem are you solving?
00:17:06.640 --> 00:17:08.720
So maybe I don't need to use AI.
00:17:08.799 --> 00:17:15.200
And the second thing he says, and and and and companies that you're training should also think about this: what makes me or what makes us special?
00:17:15.440 --> 00:17:16.799
Why would someone care about this?
00:17:16.880 --> 00:17:17.440
Is it news?
00:17:17.519 --> 00:17:18.160
Is it information?
00:17:18.319 --> 00:17:20.240
Is it a genuinely interesting story?
00:17:20.400 --> 00:17:24.160
And when you can ask those two questions, then you can say, Well, should I use AI?
00:17:24.240 --> 00:17:25.519
And then how do I use AI?
00:17:25.680 --> 00:17:31.279
So those two questions for both sides, the the journalist and the comms professional, how important do you think they are?
00:17:31.599 --> 00:17:40.559
I think it's extremely important to, and more important than ever, actually, to really get into your client or your reader's shoes.
00:17:40.720 --> 00:17:51.599
Um, because fundamentally, you know, we're talking about super sophisticated machines here, and actually this is we've got to lean into our value as the humans actually in this world now.
00:17:51.680 --> 00:17:52.880
And what is that value?
00:17:53.119 --> 00:17:54.960
This crash has kept me up at night, you know.
00:17:55.279 --> 00:17:56.960
Like, what's it mean to be human now?
00:17:57.119 --> 00:17:59.279
Like, what what what do we bring to the table?
00:17:59.440 --> 00:18:00.880
What makes us valuable?
00:18:01.200 --> 00:18:04.160
Well, I get asked all the time, you know, Andrew, will AI take my job?
00:18:04.240 --> 00:18:06.799
And I've worked out a narrative to to push back on that.
00:18:07.039 --> 00:18:18.240
Look at the job you do, look at the job I do, think listeners, think of the job you do, and slice it into a series of tasks, the things that you do every day, and then look at the tasks that you can automate and the things you don't like doing.
00:18:18.400 --> 00:18:21.440
And I actually said this to a group of hairdressers from Weller in January.
00:18:21.519 --> 00:18:24.559
I said, focus on what you love and automate the rest.
00:18:24.720 --> 00:18:31.440
And what you love is what makes you human, what gives you that empathy, what really makes you chase that story or be the storyteller.
00:18:31.599 --> 00:18:32.720
The rest can be automated.
00:18:32.880 --> 00:18:40.000
So, to your point, what do you think on both sides of journalists and comms professional, what's the stuff that they should keep because they love doing it?
00:18:40.240 --> 00:19:00.000
And I do think it comes back to relationships, and and that's what when I started in journalism, it was all about developing these relationships, PR and journalist relationships, which you kind of lost over the years because of technology, the I, you know, because of the emails and the and the ways that we communicate now on social media or LinkedIn or whatever it is.
00:19:00.160 --> 00:19:09.680
And I I think it's that intuition, empathy, uh, it's that ability to sit across a table from somebody and notice a shift in their tone or their body language.
00:19:09.920 --> 00:19:18.240
Often I've just felt like something's off, or actually there's a there's a I've had an idea, I've been out and about talking, it's often in conversation with somebody.
00:19:18.400 --> 00:19:21.200
And that is a specific, it's human conversation.
00:19:21.279 --> 00:19:34.319
It is not a conversation with an AI that necess that will spark the best ideas and investigations and campaigns as well, because it's really putting yourself in the reader's shoes and thinking, is this interesting?
00:19:34.480 --> 00:19:35.759
What would I want to know?
00:19:35.920 --> 00:19:36.960
What did they want to know?
00:19:37.119 --> 00:19:38.319
What's going to resonate?
00:19:38.480 --> 00:19:40.400
And I just don't think an AI can do that.
00:19:40.640 --> 00:19:43.119
And there's those relationships, go back to Andrew Hill.
00:19:43.200 --> 00:19:45.759
Uh he was referred to me by my editor actually.
00:19:45.839 --> 00:19:50.319
They were met at a book fair, and he said, Any examples of business books that are do something different?
00:19:50.480 --> 00:19:53.519
And uh my my editor said, Yeah, you should talk to Andrew Grill.
00:19:54.160 --> 00:20:02.559
And actually, uh Andrew and I were gonna uh have a video call, and I worked out I was at a conference that was literally one street away from the FT offices.
00:20:02.640 --> 00:20:03.440
So let's meet.
00:20:03.599 --> 00:20:07.599
We spent 50 minutes together, and I'm not sure how many people get to spend 50 minutes with the FT.
00:20:07.680 --> 00:20:08.720
We had a great discussion.
00:20:08.880 --> 00:20:12.480
I talked about his book, I talked about what I was doing, he asked me a bunch of questions.
00:20:12.640 --> 00:20:18.000
I think we built a relationship, and he also knew about me because he was sent my book years ago when it came out.
00:20:18.079 --> 00:20:23.440
So I think that's so important investing the time because you can't replicate that with AI.
00:20:23.599 --> 00:20:32.480
Um, and I hit one other thing he said, uh, which I anyone, anyone who is an author or an aspiring author, he said, your book is the heaviest business card you'll ever have.
00:20:32.640 --> 00:20:33.759
And he was so right.
00:20:33.839 --> 00:20:36.799
Um, the book that I gave you, I'm sure you've spoken to people about it.
00:20:36.880 --> 00:20:39.200
You thankfully left a review on Amazon.
00:20:39.279 --> 00:20:43.279
Um, that is a calling card that that is a relationship that you can't replace.
00:20:43.519 --> 00:20:46.400
And the more I experiment with AI, the more I crave books.
00:20:46.640 --> 00:20:49.519
I think it's really I've craved that printed page.
00:20:49.759 --> 00:20:53.359
I've been reading more than ever, and yeah, your book's just behind me now.
00:20:53.440 --> 00:20:55.599
Uh on my my shelf of AI books.
00:20:55.920 --> 00:20:59.359
So, how has your own view of what's uniquely human in journalism evolved?
00:20:59.519 --> 00:21:03.759
Is there anything you now happily outsourced to AI you once considered a core human skill?
00:21:04.079 --> 00:21:11.920
This is going to sound probably a bit odd to some people, but I use an AI tool called Whisperflow and I voice dictate a lot.
00:21:12.000 --> 00:21:21.440
And I used to think that that messy first draft as a journalist was where all my, you know, the the blank page was where all my thinking took place.
00:21:21.599 --> 00:21:33.200
But now I do work very differently in that I will brain dump my ideas from my head onto a blank sheet of paper using Whisperflow, just dictating.
00:21:33.359 --> 00:21:39.279
And then often I will go away, do something else, cut and then come back to it and sort it out from there.
00:21:39.440 --> 00:21:43.599
Because and then I find that AI can really help me structure my thinking.
00:21:43.759 --> 00:21:47.200
My brain is the type of brain that shoots off in a million directions at the same time.
00:21:47.440 --> 00:21:52.799
And I have lots and lots of different ideas, and I often can't get them out all at once and coherently.
00:21:52.880 --> 00:21:55.920
So I feel like it's been really helpful for that.
00:21:56.079 --> 00:21:59.680
And I know other writers would say differently, they would say, you know, I can't, I couldn't possibly.
00:22:00.160 --> 00:22:00.799
Work without that.
00:22:00.960 --> 00:22:09.359
The bat the wrestling with the blank page and the writing is the thinking, which I agree that writing is thinking, but you can do that at different stages.
00:22:09.519 --> 00:22:16.160
So that's how uh yeah, I never thought I would be saying, Oh, I use an AI dictation tool to structure my thinking.
00:22:16.559 --> 00:22:17.519
You would have read in the book.
00:22:17.599 --> 00:22:19.839
I started my book Journey back in 2009.
00:22:19.920 --> 00:22:21.440
I wrote a book called Twitter for Business.
00:22:21.519 --> 00:22:24.319
And I push back for years and years because I had a problem with that blank canvas.
00:22:24.559 --> 00:22:25.839
I didn't know how to start.
00:22:26.000 --> 00:22:30.160
And my agent Michael Levy about seven years ago said, Why don't you start a podcast?
00:22:30.319 --> 00:22:33.359
And then maybe the guests on the podcast can actually make it into the book.
00:22:33.519 --> 00:22:39.920
But what I did with those 60 interviews, I had about half a million words to play with that were all transcribed.
00:22:40.079 --> 00:22:43.279
So my thing was I didn't have a blank page, I had a very messy page.
00:22:43.359 --> 00:22:48.240
And then what I could do was structure that to say, here's why I think I should be curious about this.
00:22:48.400 --> 00:22:51.759
Here's an expert, here's some commentary, and here's what I think as well.
00:22:52.000 --> 00:23:00.160
And actually that helped me get the book out really quickly, but I had to use AI because having 60 hours of podcasts, there's no way I could have gone through all that.
00:23:00.240 --> 00:23:10.400
So in a way, we're quite similar, but I needed the words on the page to then move them around, like a jigsaw puzzle you throw on the ground, wear all the blue sky, and then move it all around to you have the blue sky there.
00:23:10.640 --> 00:23:12.640
And it's quite a fun way of working, I find.
00:23:12.799 --> 00:23:15.440
Like I and I I like to just let it settle as well.
00:23:15.599 --> 00:23:23.359
I'll come back to it and come back to it, but AI really helps me get all my thoughts out there and my ideas out there in a way that I think I couldn't before.
00:23:23.680 --> 00:23:30.880
So you talked about the problem with fake experts, and I'm seeing more and more of that, and and hopefully you can weed them out and journalists can allow to weed them out.
00:23:30.960 --> 00:23:33.599
But do you think audiences still care whether a piece was touched by AI?
00:23:33.759 --> 00:23:37.119
Or is the new standard simply, can I trust this and is it genuinely useful?
00:23:37.359 --> 00:23:38.720
Well, they would like to know how it's been used.
00:23:38.799 --> 00:23:42.559
So research, you know, digging into data, telling better stories, that's all great.
00:23:42.720 --> 00:23:44.720
I think they just want transparency.
00:23:44.960 --> 00:23:53.279
So it is just a transparency issue, and where I do see problems arising is more AI images in journalism.
00:23:53.359 --> 00:23:55.680
So you want to be very careful about use of that.
00:23:55.839 --> 00:23:59.759
They want to know that they're reading genuine, you know, human analysis and thoughts.
00:23:59.920 --> 00:24:02.480
The New York Times actually got rid of one of its freelancers.
00:24:02.559 --> 00:24:12.319
Um, you might have seen the story, which um lifted a re book review from The Guardian, had used AI essentially, didn't realise that AI was spitting out essentially the same review.
00:24:12.480 --> 00:24:14.720
They want to know they're reading authentic work.
00:24:15.039 --> 00:24:18.000
So let's look at lessons for curious professionals beyond journalism.
00:24:18.160 --> 00:24:27.440
For someone listening who was not a journalist but works under pressure, a lawyer, consultant, or executive, what can they borrow from your journey that would help them go from deadlines to data in their own world?
00:24:27.680 --> 00:24:35.440
So where I see the power with AI, whatever profession you're in, is the combination of AI and human expertise.
00:24:35.599 --> 00:24:37.680
So not to lose sight of that.
00:24:37.920 --> 00:24:57.680
And whatever your skill set is, to lean more heavily into that and then apply AI to the process and really to have your brain engaged at all stages is tempting because AI is obviously so fast now that to offload your thinking, and that is a really really dangerous territory.
00:24:58.079 --> 00:25:04.640
So if you were editing a front page splash about your own journey into AI, what would the headline and standfirst say today?
00:25:04.720 --> 00:25:07.279
And what do you hope they will say in five years from now?
00:25:07.680 --> 00:25:14.160
I'm gonna go with the silly, a silly headline um for something like a tabloid, maybe like hack to the future, something like that.
00:25:14.559 --> 00:25:15.839
Oh, I see what you did there.
00:25:15.920 --> 00:25:17.200
Oh, that was obvious.
00:25:18.000 --> 00:25:19.680
Or prompt and circumstance.
00:25:19.839 --> 00:25:20.400
I don't know.
00:25:20.480 --> 00:25:21.759
Um, from print to prompt.
00:25:21.920 --> 00:25:28.319
I'm not naive to the fact that there are so many issues around AI, but I'm I also think there's so much potential.
00:25:29.359 --> 00:25:32.480
You know, the and that's where we but we need to lean into the potential.
00:25:32.799 --> 00:25:36.000
So I couldn't leave without asking you about the next generation of journalists.
00:25:36.079 --> 00:25:44.000
So I see this all the time, and maybe journalists is a little bit more immune to it at the moment, but someone coming to the workforce fresh, they've now got all these tools in front of them.
00:25:44.079 --> 00:25:50.160
What I'm finding is either the jobs are not available for early entry people, or they're saying we need to do more with that.
00:25:50.400 --> 00:26:00.799
The podcast I did a couple of weeks ago with Tim Cook, episode two of the current series, he is actually an elementary school teacher, and I'd recommend people listen to this after they've listened to this episode.
00:26:00.960 --> 00:26:06.079
And he sees kids actually, the difference between them leaning on AI tools versus not.
00:26:06.240 --> 00:26:18.319
And he pointed to some research in psychology today that said, if you're 45 or older and you lean on AI, it's actually okay because you've done it before, back to the journalism skills, you know how to hunt out a story.
00:26:18.480 --> 00:26:25.839
But if you're under 25, you're really at risk because if you're leaning on AI tools for doing your job and we turn off the AI, can you still do that?
00:26:26.000 --> 00:26:27.119
Have you learned how to do that?
00:26:27.279 --> 00:26:29.039
Have you engaged your critical thinking?
00:26:29.200 --> 00:26:34.640
What would be your message to young professionals coming into the industry and how do they gain the skills?
00:26:34.720 --> 00:26:37.599
And how do we ensure that there are still places for young journalists?
00:26:38.000 --> 00:26:47.200
Is something actually that really concerns me at the moment, and actually, one of my next newsletters or an AI for media newsletter on LinkedIn, I'm going to write about that very topic.
00:26:47.359 --> 00:26:56.319
But I've been thinking about this, and my advice would be flip your AI use if you can to asking AI to interrogate you.
00:26:56.480 --> 00:26:58.079
Like get it to ask the questions.
00:26:58.160 --> 00:27:01.680
Like you said, like AI is very good at asking, it likes questions.
00:27:01.920 --> 00:27:03.839
So get it to ask you questions.
00:27:04.000 --> 00:27:21.599
Start with like what your outline story is and say, like, right, now question me so we can get to the heart of what and and then you can help your expand your ideas, maybe what your angle is, maybe what the story actually is, who you should speak to, all these things.
00:27:21.759 --> 00:27:32.480
Rather than letting, I just think don't let AI do the thinking for you, make it refine and act as a thought catalyst to be able to actually spark that process.
00:27:32.799 --> 00:27:42.319
I use it as a decision partner, and I was recently submitting a proposal to a client, and the AI asked me, Would you like me to show you how this proposal will land emotionally?
00:27:42.480 --> 00:27:45.440
Because it knew that the client was also a friend of 15 years.
00:27:45.599 --> 00:27:46.720
And I hadn't thought about that.
00:27:46.880 --> 00:27:50.559
It then explained that and I looked at that and I thought, well, oh, that will land well, that will land well.
00:27:50.640 --> 00:27:54.799
But I hadn't even thought about that treatment in all of my years in consulting.
00:27:54.960 --> 00:28:01.200
I never thought about the emotional impact of what you're delivering, especially when someone that is the client knows you very well.
00:28:01.519 --> 00:28:10.799
And I think this is where AI and where I see its power, particularly having worked as a freelance journalist for the nationals for a long time in my career.
00:28:10.960 --> 00:28:30.799
I it's like, you know, I found like I had this, I had this, as you say, this thought partner, like somebody I could spar with, somebody, uh, an AI I could actually have this back and forth with that would unpick and help me uncover more of my own thinking in and I love that, that you can actually expand what you're capable of.
00:28:31.039 --> 00:28:40.640
And because I put all of my work predominantly in one AI, it's got a memory there, whether it be in a client space or with in the threads that I use, it picks up things.
00:28:40.720 --> 00:28:48.559
It says, because you're the actual futurist, it is if I have this long-term 2IC that knows everything about me and it just surprises me every single day.
00:28:48.799 --> 00:28:53.119
Yeah, it's the memory thing is quite when that arrived, I think that's changed everything.
00:28:53.200 --> 00:28:59.920
And and it is uncomfortable, but also, yeah, the power of it to be able to push you into corners you might not have gone to.
00:29:00.240 --> 00:29:05.039
So we're almost out of time and we're up to my favourite part of the show, the quick fire round where we learn more about our guests.
00:29:05.200 --> 00:29:05.920
Window or aisle?
00:29:06.160 --> 00:29:06.480
Window.
00:29:06.640 --> 00:29:08.240
Like to look out the window, then go to sleep.
00:29:08.480 --> 00:29:09.359
iPhone or Android?
00:29:09.599 --> 00:29:10.559
iPhone, always.
00:29:10.880 --> 00:29:12.559
Your biggest hope for this year and next?
00:29:12.799 --> 00:29:17.519
That people lean into their human skills and uncover their own unique value.
00:29:17.759 --> 00:29:21.839
I wish that AI could do all of my cleaning, cooking, everything around the house.
00:29:22.160 --> 00:29:23.920
The first thing I asked ChatGPT.
00:29:24.160 --> 00:29:29.519
I asked it for i consumer finance angles for a pension campaign I was working on.
00:29:29.759 --> 00:29:32.319
Is generative AI evolution or revolution?
00:29:32.640 --> 00:29:34.079
Day by day it's an evolution.
00:29:34.160 --> 00:29:35.680
Ultimately, it's going to be a revolution.
00:29:36.000 --> 00:29:37.279
The app you use most on your phone.
00:29:37.519 --> 00:29:37.920
WhatsApp.
00:29:38.240 --> 00:29:40.319
How do you stay digitally curious?
00:29:40.640 --> 00:29:44.880
Constant experimentation and being optimistic.
00:29:45.119 --> 00:29:46.960
What's the best piece of advice you've ever received?
00:29:47.119 --> 00:29:55.920
My mum actually wrote The Only Time We Have Is Now and stuck it on a piece of paper on the fridge and she died two weeks later, which is was devastating, but also I've never forgotten that.
00:29:56.000 --> 00:29:57.519
And it's great life advice.
00:29:57.839 --> 00:29:59.039
What are you reading at the moment?
00:29:59.279 --> 00:30:04.960
I am reading uh The Infinity Machine, Demis Sabis Deep Mind and the Quest for Super Intelligence.
00:30:05.119 --> 00:30:09.839
I bought it the other day, it's quite similar to super intelligence, but it's absolutely brilliant.
00:30:09.920 --> 00:30:10.720
I really recommend it.
00:30:10.880 --> 00:30:12.160
Really well written, super interesting.
00:30:12.400 --> 00:30:14.160
Who should I invite next onto the podcast?
00:30:14.480 --> 00:30:15.519
Nick Newman.
00:30:15.839 --> 00:30:16.720
He's super interesting.
00:30:16.799 --> 00:30:25.839
He's a senior research associate at Reuters Institute for the Study of Journalism, and he's got some really interesting views of how AI is changing news consumption.
00:30:25.920 --> 00:30:28.960
And so how what we're going to take in in this new world.
00:30:29.200 --> 00:30:30.400
How do you want to be remembered?
00:30:30.640 --> 00:30:35.119
As someone who helps people be brave enough to embrace change and take risks.
00:30:35.440 --> 00:30:41.680
Harriet, what three actionable things should our audience do today to better understand how we can use AI for media and corporates?
00:30:42.000 --> 00:30:45.440
First, um, go deep and narrow rather than wide and shallow.
00:30:45.519 --> 00:30:56.480
So choose something specific and really whether that's research and analysis, um preparing interview questions, whatever it is, and really going deep on how AI can help with that process and refining it.
00:30:56.640 --> 00:31:06.240
Second, learn the boundaries, so test where it fails, ask it for sources, check quotes, put it under pressure, find the gaps so you really understand it.
00:31:06.319 --> 00:31:10.480
And third, I'd say build the habit of human review into your process.
00:31:10.640 --> 00:31:20.720
So use AI as a thought partner, research partner, thinking partner, but keep a human in charge, your editorial judgment in charge at all parts of the process.
00:31:21.039 --> 00:31:25.519
Harriet, I knew when we met at News Rewide, you're a fascinating person with great insight.
00:31:25.839 --> 00:31:29.359
Even more delighted to have you on the podcast and unpack some of those things.
00:31:29.519 --> 00:31:31.599
But how can we find out more about you and your work?
00:31:31.839 --> 00:31:33.519
Search for me on LinkedIn, Harriet Mayer.
00:31:33.680 --> 00:31:40.079
I write an AI for media newsletter and I do lots of speaking and training and all sorts of things.
00:31:40.319 --> 00:31:41.279
Harriet, a pleasure.
00:31:41.359 --> 00:31:42.559
Thank you so much for your time today.
00:31:42.720 --> 00:31:43.359
Thank you very much.
00:31:43.519 --> 00:31:44.640
I really enjoyed that.
00:31:44.880 --> 00:31:47.599
Thank you for listening to Digitally Curious.
00:31:47.759 --> 00:31:52.240
You can find all of our previous shows at digitallycurious.ai.
00:31:52.799 --> 00:31:57.279
Until next time, we invite you to stay digitally curious.