The AI Ambition Execution Gap

Chris Norton 0:00

Sixty-four percent of CEOs are actively championing AI in their organization, but only thirteen percent of them are actively using it in their organization, and seventy per cent of junior team members are running the tools every day with no oversight, no guardrails, and nobody checking their output. What could possibly go wrong? I'm Chris Norton, and this is Embracing Marketing Mistakes, the show that teaches you growth from failure. We are still live from the AI for PR conference on the South Bank of London, and we've got two more conversations with one strong question running all the way through it. What is the gap between AI ambition and AI execution? Stuart Bruce is back on the show, my old boss, and um I managed to grab him and dive into the Copilot and some of the top tips and tricks that you can use for the Copilot, because you heard a lot of people. So, as always, relax, and let's get into AI, PR, and marketing. Enjoy.

Catching Up With Stuart Bruce

Chris Norton 1:11

Right, Stuart, Bruce, welcome back to the show.

Stuart Bruce 1:14

It's a pleasure to be here. In this brief I'm you're not even on the speaking list today.

Chris Norton 1:19

What's happening, Stuart? You used to be on every speaking.

Stuart Bruce 1:20

I know, I've been I've been shunned. It's all over. Uh no, it's actually really, really good fun being actually being at a conference where you don't need to kind of think about, oh God, what am I gonna say? Have I gotta prepare? Yeah. And just be able to sit down and absorb stuff.

Chris Norton 1:35

Yeah, yeah. Have you read the book then, or have you are you going to read the book?

Stuart Bruce 1:39

Um, I have the book and I have read some of the chapters. Okay. Um, but my I despite the fact I ordered it ages ago. My physical copy only arrived um uh this week. Oh uh I've had my digital one for over a month. But I don't really I prefer to read the paper one, so that's the reason I've not read it.

Chris Norton 1:56

I'm with you on that, because that you can pick it up with books like that, like you like the social media handbook one we wrote ages ago. When you it was you need to like do them in bits. You can't they're not like a it's not a gripping story, is it?

Stuart Bruce 2:06

No, exactly. It's so yeah, funny, I I I read fiction on my Kindle, but work-related books I prefer paper.

Chris Norton 2:12

Yeah, me too, me too. So um so what have you been up to then? We haven't seen you for about a year on the podcast. What have you been doing?

Stuart Bruce 2:20

Mega busy, and most so last year was all about AI adoption. Basically, in-house teams, agencies use AI. Yeah, this year it's all about AI as a stakeholder or generative engine optimization if you want to take a simplistic view of it. Um, and it's quite a controversial position about which of the two it is.

Chris Norton 2:41

Yeah, I know uh well um I interviewed, I've I've spoken to Johnny on this podcast, and um he said he didn't care which one it was, AEO, GEO. But what I did like in just before we came down here, Stuart, is that you told me that I was best practice. Uh not me personally, the the agent prohibition have got a good GEO, which is which is good because that's what we do, isn't it? And it's what you do. Yep. So and I I like the fact that you used as an example of best practice, that's good. So thanks for thanks for doing that.

Stuart Bruce 3:08

But I know, but it's true. It's it we when I was kind of doing my experiments, it were just kind of but we're actually only half surprised because you've always been good at SEO.

Chris Norton 3:16

I'd love to know what you picked out as holes on what we do, and then you can tell me, and I'll get that re-optimised. Because it's an ongoing thing, yeah. It's like painting the uh Brooklyn Bridge. Is it the Brooklyn Bridge?

Stuart Bruce 3:26

Yeah, Brooklyn Bridge or the Forth Road Bridge. Forth Road Bridge, that's it. Um but the thing is, I mean, it I think there's a bit of cobbler's children about it, because we've been doing so busy doing this type of stuff with clients that our own website is far from perfect.

Chris Norton 3:41

It's not that bad though. I read your newsletter.

Stuart Bruce 3:43

Oh, yeah, no, the newsletter works well, and the newsletter's really good. Um, but the actual website is, you know, we've only just added schema to it. Yeah. Um there's a whole bunch of other things that could be um kind of put on

Why Copilot Wins At Work

Stuart Bruce 3:56

there.

Chris Norton 3:56

I need to update my schema. It's uh pain in the ass, but it's it works, doesn't it? Um, yeah, so we were just before we came down here, you were talking about Copilot, and you had a rant about Copilot. Oh yes. So I'm I'm I'm intrigued because I've got a view on it. So give me your rant.

Stuart Bruce 4:11

Well, I can rant away. So we've had a couple of at least two people on stage, probably more, kind of making disparaging remarks about it. Okay. Um, you know, to be blunt, that probably says more about them than it does about Copilot. Because the reality is, in a corporate environment, that's what people are gonna uh need to use. You can't be at the cutting edge. That's just not got it's just not an option. And if you actually spend enough time learning how to use Copilot properly and kind of push it to its parameters of what it can do, it simply isn't true that it's kind of not as good. In fact, for a lot of things, it's a lot better. So Charles. Well, the reason it's a lot better is because is because what Microsoft called work IQ, and that basically means it's grounded in your data. Context. Yeah, context. Yeah. So if you say, and and the best example of that is if you compare using Claude standalone or Claude within Copilot, the answer of Claude within Copilot is mind-blowingly better because when you ask it to go off and do a research project or something, it already knows the types of things you're interested in. So it actually research, without you telling it, it researches the stuff that's relevant to you. So you know, if you were doing it in another tool externally, or even if you were somebody using something like Perplexity, you'd probably have follow-up questions, or you'd have to try harder to write a detailed prompt. But if it already knows everything about you, you don't need to.

Chris Norton 5:33

So are you using Co-work inside Copilot?

Stuart Bruce 5:36

So so well, two things.

Chris Norton 5:37

So brand new that, isn't it?

Chris Norton 5:38

Stuart Bruce 5:38

Yeah. So that little well, we've been using it for about three months. Right, yeah. Uh, because it's because you can have there's something called Frontier Program. It's like a buying into beta, isn't it?

Chris Norton 5:47

Yeah.

Stuart Bruce 5:47

Um so we signed up for that three months ago, so we've been using it quite extensively.

Chris Norton 5:51

But isn't it really expensive?

Stuart Bruce 5:53

Well, in the Frontier program, it was free. Um, and for those of us that joined it, it's still free till the first of July. So I'm about to get my money's worth of before.

Chris Norton 6:01

I want to know how many tokens you're using, because that's that's the issue. Because it just it there was there was somebody on there was a couple of people that did say some interesting things on stage today. Um, because someone mentioned loops, which is the big thing in in California at the moment. These loops where you use loops to iteratively improve your prompts.

Stuart Bruce 6:17

Yeah.

Chris Norton 6:17

So it's literally improving the conversation and it's improv it's self-improvement, which some would say is is is moving towards AGI because it's improving itself. Now that I find that interesting, and you can do that in Co-work, and in theory you can do it, but I've heard that the pay, the price I know that's this because I got it from my IT people on Thursday. The pricing model of Copilot is on tokens, and it's predicted tokens. They don't know.

Stuart Bruce 6:44

So there's a pay as you go. There's a pay as you go model. Yeah. So, and you can set caps by user, by by group, or by tenant. Yeah. So you're not going to accidentally spend money. Um and I um then Microsoft created a spreadsheet, which, if anything, I think overestimates what it's gonna cost. Because realistically, not everybody in a company or a team is a heavy user or a cutting-edge user. So most are gonna have a lower level of use than Microsoft's spreadsheet. Yeah. You know, and the spreadsheet just you know it shows it's gonna be something between 85 to 140 quid a month per user. Um

Tokens, Costs And Co-Worker Automation

Stuart Bruce 7:25

for what it can do, that's actually not that expensive. Uh I mean give one practical example. We ran a 85, 85 to replay that. 85 to 145 pounds a month for the tokens.

Chris Norton 7:37

That's such a big gap, though.

Stuart Bruce 7:39

Yeah, because it's the £145 is for a really heavy user. Like me. Uh yeah, they well s possibly not you might be even higher than that, so I might be even higher, but for your average heavy user. Right. Um, but just to give one example of how that would, you know, get its money back. Um we organized an invite-only breakfast recently with Simon Boor, uh kind of former kind of chief executive of the communications service. And we sent out the invites in HubSpot so we could track the opens and everything. As the replies came back in, Co-work, um, co-pilot Co-work intercepted them. It filed them into one of three folders, kind of yes, no, um, out of office so they could get a reminder, which it did. Um, it then added all the names to the spreadsheets. If there wasn't, if there were things like a job title missing, it would go off and research the job title and populate that column in the spreadsheet. I love Co-work. Um then it would it uh it generated the name, but it basically did all of the admin tasks for the breakfast. Now, if we'd high if you know, we we sometimes use a kind of virtual PA, you know, if if we'd have been paying them to do it, that would have been several hours' work. Um unless they were using Yeah, unless they were using now the the the big the difference between the two is if the virtual PA you wouldn't have needed to check their work. Uh because we you know you you trust them, you know them, so you you you you don't need to check it. But then again, if they were new, you wouldn't. Yeah, you wouldn't be able to do that, you would have to check it. Um whereas with the Co-work stuff you did. Yeah. Um and it was kind of, I would have said it was 85, 90% accurate. Um the you know, and sometimes it confused you by what it did. You know, so for example, one person replied to say they couldn't do it, but could they send a colleague? Um and it drafted an email uh kind of accepting it, and um and it knew that. So in the notes field, it's it it kind of said that's why this person was on the list, even though they weren't invited. Amazing. Um and because some of the integrations don't work, it couldn't, it it tried to uh it tried to update HubSpot, but because we hadn't given it access, MCP access. So it actually said you need to update HubSpot.

Chris Norton 9:54

It does that to me. I'd say what's weird, I mean we're talking about Copilot here, but Claude, I said I think I said this to Andrew, is the way it talks to me. Uh I I said something the other week to her and it went on it, mate. Because what did else it said on it, mate, uh, and it said it it talks in like in you I don't know if it's mirroring your language and how I'd speak.

Stuart Bruce 10:16

Well and that's but that's the other thing. So so I've the model I use most is Claude within Copilot.

Chris Norton 10:21

Yeah. I'll be honest, I haven't used so I've got Copilot, I've got Gemini, I've got Co-work, I've got Claude, obviously, um, and a bit of perplexity, and I've tested I've tested them all for different things, research. I wanted Co-work, I I now feel like Co-work's going to be a waste in Copilot, because I've got Co-work in the other. Have you got scheduled tasks in co-pilot Co-work?

Stuart Bruce 10:44

Um yes. Right. So that's so that breakfast example, um, then that was how that worked, is every morning it would kind of do a sweep of the emails that had come in. Yeah. Uh, and then do the replies and everything. Yeah, it's

How To Get More From Copilot

Stuart Bruce 10:55

Chris Norton 10:55

amazing, isn't it? Um, so what are your hot tips on co-pilot then to how to get the most out of co-pilot? Because there'll be people out there, right, that have got like marketing people that have they've been given it, they've been given a license, they've you they probably use it on the meeting note, yeah, and they probably use it like a chat box.

Stuart Bruce 11:13

Yeah, we we have clients at the moment where they've been given licenses. They might have been given training on how to use Copilot, but because it's not embedded in comms and corporate affairs and communications, sometimes they struggle to see how you can make it relevant. Um so the the way to approach it is to do the list of things you want it to do, um, and then work out whether that's possible to get it to do it. Because very often the answer is yes.

Chris Norton 11:39

Yeah, yeah. I know. I well, because I I that guy I was telling you about, who Sean remained nameless on this podcast, um, told me that Copilot was just chat GPT light, and it isn't as good. And I said, I I'd beg to differ then, mate. Because I said I I think some of the you maybe the results, I mean comparing the models, right? Like Mythos or Fable now that we haven't got that I had for 48 hours and is now gone again. Um the my my my issue is the worst models that we've got now are are pretty pretty good.

Stuart Bruce 12:13

100%. So I get really frustrated by people chasing the late, I mean and Gregory talked about it on the panel this morning. The latest shiny tools, they really don't matter. 18-90% of PR people aren't even touching the edges of what a tool a year old could do. No, true. Um so the idea that you need to do this cutting edge stuff is ridiculous. You you basically, is it going to have an impact on clients? Is it or if you're in-house, is it going to have an impact on your kind of internal stakeholders and colleagues? Um and you don't need the latest shiny tools to do that. And the thing about Copilot is some of the things that people criticize it for are actually its strengths. Because the reason the guardrails are there is if you are working for a pharma company or government or financial services, you need them. And it and it basically takes away the worry. Yeah. You know, you'd you you can stop thinking about those things as much because it's doing it for you.

Chris Norton 13:05

And and you might have used it two years ago, right? You should have got this episode sponsored by Copilot. Um, and you might have used it like two years ago and thought, oh, it's pretty crap. Whereas in actual fact, um it's the latest version of Chat, it's about a day behind ChatGPT, they update it, and so it's five point, we have five or six. Uh exactly, like you said.

Stuart Bruce 13:24

Yeah, but they they roll the when then now when a new model comes out, um, it gets incorporated the same day.

Chris Norton 13:30

But that's why that's why Microsoft have moved to to work with co-cloud as well. They're diversifying their AIs because um, and this is the rumour that that Microsoft are working on their own lab because they're making so much money out of it. It'll be it's gonna be very interesting.

Stuart Bruce 13:45

Well, at the build at the build conference, they launched a whole range of kind of their own AI models. Um there's also talk of using um their own version of DeepSeek um for to power co Copilot Co-work. That worries me that. But it doesn't because it the whole point is because it's open source, if you're running it on your own servers, you know, there's no reason for it to. You know, they're basically just taking it and kind of running their own version of it. So it means they can do so cheaper. But also the the token cost of Co-work in Copilot is cheaper than Claude's token costs. Um and it was a bit it was one of these kind of statements in the launch where you had to kind of read it about five times to understand what it meant. But but crudely, it's for two reasons is one, Microsoft are buying it at wholesale cost, so you know they're passing on some of that discount to customers. Um, but the other one is it's not just pure Claude that you're using, it's it is optimized for co-pilot and mixed. Yeah.

Chris Norton 14:47

So they're it's more efficient. Yeah, it's in your environment and it can see your stuff. Can't that and that is the like you said, the only thing I I I do have some reservations with with Copilot, because for instance, I'll ask it to find me because it sounds brilliant. They go, you can search your emails, you can search your folders, you can which you can do in Co-work as well, by the way. Claude the uh original. Um but sometimes I find its search is a bit shit. I don't find it's as good as it's is meant to be, like it'll it'll look for things and doesn't quite it doesn't quite find it and put it together the way if you did research for yourself.

Stuart Bruce 15:21

It can be a bit of a hit and miss. Yeah. Um but uh that's partially because those people have built really bad historic SharePoint sites. Um so we're just doing a project at the moment where a client has migrated from Dropbox to share SharePoint. So we're expecting that to work far better because ever not it's not quite perfect, because you couldn't get the migration perfect, but it is better than what a lot of people have. I mean our own version, because it's our SharePoint dates back to 2011, so to say that it is a mess is the understatement of the century. So it's perhaps not surprising that Copilot sometimes struggles to make sense of it. Um you know, and there are things we could do to tidy it up. And interestingly, now we've got Co-work. You might actually be able to use Co-work to tidy it up.

Chris Norton 16:14

Oh yeah. Yeah. Well, I'll tell you what I did straight away first day I had it was um remove all duplicates. I'd done some duplicate, and there's about 40 gigs worth of duplicate images and all sorts on the machine. It instantly removed them all and sorted out and reduced my um capacity on my laptop and it helps, you know, it's good.

Stuart Bruce 16:32

Well, while we've still got the free version of because we were the frontier model. So before the 1st of July, that my intention is to try to make a better version of SharePoint.

Chris Norton 16:40

It's like your own version of Independence Day. Uh the 4th of July, wasn't it? So there's a bit of a rush to get it done. So your hot tip then for if marketing people in-house, if they want to get the most out of Copilot, how do they do that?

Stuart Bruce 16:53

First, be crystal clear about what it is they want, but also that there's no kind of once it's a bit like kind of what do we call it, uh Dave Beresford and kind of the minimal minimal gains. Yeah, marginal gains. Marginal gains, yeah. Um there's 101 small things that once you've done them, you get much better results. Like the number of times you go in and look at people's co-pilot setup and they haven't personalised it, you know, there's a whole section where you can tell it what type of answer you want, you can teach it your writing style, and the best way to do that is actually to use an agent to analyse your writing. So you don't have to tell it what style you want to get it to understand your style and then write instructions for it. But you go in and people haven't done that. Um the frustrating thing is it's part of the onboarding. You know, you get told to do it, but people just skip past it and don't do it. Um so yeah, that that will be the it's it's it's 101 little things. It's not there's not kind of a big magic wand.

Chris Norton 17:47

But it'll certainly so I asked I asked my um we've had it for well, we've had numbers we had all the tools paid for separately, and then we we've embedded it in the agency. So everybody in my agency, pretty much, I think 90%, I didn't push it on anyone. I was asking them what they wanted, what they wanted to use, and then most of you Copilot, then when I did a load of training, I trained the team up, and I did a sort of a straw poll the other day, and pretty much 90%, only one person said they weren't using it regularly, okay, they're using it daily, but most of them are using it regularly every day, which is shows you how important.

Stuart Bruce 18:21

So one of the things you can do with Copilots is you can actually go into the dashboard and see how people are using it. Yeah, that's true. Um, which is kind of really I had already done that, but I'll have to ask them as well. Um but the second thing is you've just been is this uh lots of people start using AI and companies as individuals, and the real power is when you use it as a team. So things like um in Copilot, it used to be called the Prompt Gallery, but they've just renamed it back to its own old name, which was a Prompt Lab. Uh so it used to be called Prompt Lab, then they called it Prompt Gallery, which was a good name. Yeah. Now they've got called it Prompt Lab again. Lord knows why. But people don't use it. You can so if you create prompt, you can then share them with your team.

Chris Norton 19:00

Oh, right. Because we use we I created an I created a GPT, custom GPT, whatever you call them in that. Agents, they call them in there, which is confusing because agents are something else as well. Uh anyway, I created a and I created a prompt agent so it will master your prompt for you. So you write your prompt, you put it in there, and I shared that with my team. But now you're saying there's something that then when you've got an excellent one, you can stick it in a library within your so you can all use it, which is brilliant, actually. I didn't know that. That's it.

Stuart Bruce 19:26

No, no, no. Yeah, and that's an example of people there's loads of stuff that people just don't use. Yeah, and and basically that's because Microsoft haven't done a brilliant job of communicating it. Well, it doesn't help they've got they've got like something like 70 or 80 products called Copilot.

Chris Norton 19:40

Yeah, well I that's that's what I actually thought that's what you were gonna rant about. Right, dear what's he called? It's not Bill, not Mr. Gates. Dear Microsoft, please can you remove Copilot from every single application and have it in one thing that does all of them? I don't understand why you've got to have Copilot in the code.

Stuart Bruce 19:58

That's actually what they've just announced. It's terrible. Yeah. Um well it isn't because the whole the whole point being is there there's a rationale to it, it's meant to be frictionless. Yeah. You you don't other AI tools force you to think about using AI, whereas the whole point about Copilot is you don't need to. It's there when you need it. So you don't need to come out of your workflow, you don't, it doesn't need to stop your brain. You can just it's there ready.

Chris Norton 20:23

But then the best part of it is a separate app on the desktop, the desktop that you download from Microsoft.

Stuart Bruce 20:30

Yeah, but you can do most, well, for one thing is they're now um making them the same. So it's that's not

Chris Norton 20:37

so what are they gonna do? What they they've just lit I have I've literally given them an idea and they've just turned it around before we finish this podcast.

Stuart Bruce 20:43

No, no, they've been doing that, they've they've already done that. So you've already but at the moment because of the way they roll stuff out, it's it's not synced. So because they've just launched a new user interface for the desktop or the web version, it hasn't like filtered through. Filtered through. Yeah, yeah, yeah. Um in fact, my desktop hasn't even I don't even have the new interface yet, uh, which is a bit frustrating because other people around my tenant do. You've got it. Yeah. Yeah. Uh and I'm meant to be doing training next week.

Speaker 3 21:09

What were you saying about you don't like the shiny new ties?

Stuart Bruce 21:12

Look, I I can't train people on something old, which is all I've got. Okay.

Client Expectations And The Expenses Mistake

Chris Norton 21:16

So, well, uh, I haven't prepped you on this one, but it you this show is about mistakes. Have you made any mistakes in the last couple of years since we've had anything on the show? Anything that you'd like to share that won't get you in trouble with legal?

Stuart Bruce 21:26

Always making mistakes. Let me think of a recent one. Um I can't actually think of a recent one. There must be. Um, well, uh yeah, so I I can't obviously I can't name the client. Yeah, that's fine. Um you know, but you know that situation where you have a client that kind of tells sends you an email and tells you how brilliant the work you've just presented is and uh how well your presentation um went down. Yeah. Um and then you send them the invoice and they kind of come back and say, where did all these expenses come from? But literally, the the really big bit they didn't quibble in the slightest. They thought the work was brilliant. The bit where in the contract it said we're going to charge expenses.

Chris Norton 22:04

Yeah.

Stuart Bruce 22:04

It was that is literally the only bit they're quibbling. Um I think the big mistake there is remember to tell them in advance.

Chris Norton 22:14

Manage expectations. That's the key to all client relationships. Expectation management. If you can underpromise and over-deliver.

Stuart Bruce 22:21

Yeah, you know, when I had a team, that would have happened automatically. But when it's when you were a small business, it just I remember we we talked about it, but then nobody actually did it.

Chris Norton 22:30

Yeah, it's it's like I tell you what it is like. We have you seen, I don't know if you know this, but we um we've we always we've always paid the maximum for petrol at prohibition, but they've literally with the government in the last couple of months have it upped it to 55 pence, and we were like, well, that's fine, but I don't think clients will pay that. And then we started discussing, hang on, do we even claim travel anymore off clients? Like some on projects, because we I've got a video team, they're flying around the um we do charge, you know, mileage, but then it it it's an interesting debate. Like, do you should you charge mileage if you're charging a fee for working together?

Stuart Bruce 23:03

It depends, it depends on your contract, isn't it? There's no right or wrong answer, but that's another use of co-pilot. So people forget to record mileage, you can get Copilot to go through Outlook and identify all your meetings where it looks like you've driven.

Chris Norton 23:17

No.

Stuart Bruce 23:18

Yeah. Oh, that is a hot tip. Yeah. So you identify all your so what would be the prompt, sir? Um you look literally literally just say this. Can you uh can you list all my meet uh list all my meetings where I've driven in the last six months or the last quarter, uh, and put them into an Excel spreadsheet? And it and it will do it. And then you can you can you can check that they're in your account system then. I need to do that for all the meetings that I have about your bloody AI. The reason it might not work is if you haven't put enough detail into your meeting.

Chris Norton 23:45

Yeah, yeah, yeah. I need to do that. Well, uh, that's great, Stuart. Thanks for coming on the show. Shout out to PR Andy Barr. Hi everybody, I'm PR Andy Barr. If you don't know what I do, check out this. Um anyway, um the podcast thing sort of got us behind the mic and kind of in front in front of the camera. Yeah, yeah. But it's um I'm still not used to doing, you know, like pieces to camera where you're like a BBC broadcaster.

Darryl Sparey 24:16

Yeah, yeah, yeah.

Chris Norton 24:17

Do you do much?

Darryl Sparey 24:18

Andy Barr 24:18

Andy Barr 24:18

Uh so actually my business partner, I mean I'm I'm invited to speak or get interviewed on a lot of podcasts, but my business partner Paul's got just just launched his own podcast, uh Beneath the Surface, um, which is all about what actually makes AI tick. How does it work? How does Gen AI work, all that kind of stuff? Um, released the first uh episode very recently, went really well, so you know we'll be doing more of that in the not too distant future.

Google Still Shapes AI Search

Chris Norton 24:44

Podcasting marketing is is an interesting medium.

Darryl Sparey 24:46

Oh yeah, 100%.

Chris Norton 24:47

Chris Norton 24:47

So what obviously we're at this conference today, I've spoken to quite a few of the people. What what is what you what bit you're you're doing the fighting talk section, aren't you?

Darryl Sparey 24:55

Yeah, absolutely.

Chris Norton 24:56

So are you fighting the cause for SEOs though?

Darryl Sparey 24:58

I am, I am, I know, right?

Chris Norton 25:00

Darryl Sparey 25:00

Darryl Sparey 25:00

Um boo,

Darryl Sparey 25:01

I know. Uh so I think the my premise though is uh there are uh a spoiler for what I'm gonna say later, there are five uh trillion reasons why you can't ignore Google. You know, Johnny was up on stage just now saying, like, oh, the front door of the internet is Gen AI. It's not, it's still Google, right? And even when you see Gen AI search results within um you know within the platforms, they're formulated from an internet search. So obviously Google, AI overviews, AI mode, Gemini all use Google, um, Claude uses Brave, uh, ChatGPT uses Bing. So, you know, where you rank on like traditional search engines still helps influence where you appear on um on AI search. So you you know, you can't ignore SEO and you can't pretend like it doesn't exist. It's a key input into the bigger, you know, um the bigger thing of AI search.

Chris Norton 26:02

Well, I've just literally I've got a fr I've I've got several friends who work in SEO, and uh there's a guy I know he's very good at SEO, um, and he lost about in in about November, December time, he lost about 80% of his clients, and he was getting a bit worried. He works for himself, but he's really good at what he does, so he's always busy. Yeah, um, and it his clients are all e-commerce stores, right? So they make money by search. Anyway, this thing had come out, Claude Co-work had come out, other AIs are available, yeah. But in other words, they were all saying, Oh, we can do our R SEO ourselves now. Because when I ask the if to I I don't look at I'll use some agents, I'll look at the pages, I'll assess the pages, and um, I get an excellent correct Daryl. Here's your report. These are the things that are good, these are the things that are bad. You need to optimize this and they'll change. And guess what happened? Um tanked. Yeah, four months later, 80%. Well, nine out of ten of those clients, say 90%, yeah, yeah, have all come back to him because all their stores were tanking at financially. And that's because the internet, yeah, plus people that are good at what they do. So PRs, I'm talking about digital PRs, I'm talking about traditional PRs, I'm talking about SEOs, they don't write their best techniques and tactics online and share everything.

Darryl Sparey 27:20

Yeah, yeah, yeah, yeah.

Chris Norton 27:21

Chris Norton 27:21

It's kept in a little black box because that's what makes it work. So the internet is full of all this stuff that's best practice supposedly, but actually the best practice that it's that it's learning from is out of date and isn't as good as the actual secret source. Does that make sense to you?

Darryl Sparey 27:35

Yeah, 100%. Great point. So I've got a mate uh from the Northeast, uh, he had an argument with his old man once. And his old man uh he said, you know, you're older. I won't try and do the accent. You know, you're all you're older than me, you're slower than me, you're fatter than me, or whatever you know, and you've shown me all your tricks. So I could take you in a fight. And his old man just leaned back really sagely, tapped his nose and went, You never show anyone all your tricks, son. Um and I think I think this I think exactly the same as you, right? No one's sharing, like for us, we've learned loads about um uh GEO, AI search, all the rest of it from the research reports we've done as far back as October 2024. Um but uh and we've shared loads publicly about what we've what we've learned, all the rest of it we wrote, a 3,000-word uh blog um report. Yeah, that you can.

Chris Norton 28:23

We've referenced it on the broadcast before, yeah.

Darryl Sparey 28:25

Uh yeah, yeah, bit.ly forward slash G E O for P R. Nice plug. Thank you. Uh do this for a living.

Chris Norton 28:31

Um That's what I said on my class too. Go on.

Darryl Sparey 28:35

Yeah, but uh but um but but yeah, 100%. We're not gonna tell everyone everything of what we do, and obviously we're gonna, you know. I think the other the other point that you're you're making, and we have this with clients, and I and I really want to impress this about anyone who listens to this. Um we have clients who come to us and say, we don't need you to write content for us anymore because we just get AI to do it for us. And I'm like, I can give you, you know, like case study after case study after case study of people who got sold on like by GEO bros, of you know, oh, you just talking. And talk to your agency. Yeah, exactly, right? On all this stuff, right? And oh, you can just run run your agency on autopilot, all that stuff, right? Fast forward four months, and they've had a massive Google penalty and they're wiped off, yeah, wiped off the face of the internet, basically. And and guess what? If you don't rank on Google, it's very hard to get found by by AI search. So couldn't agree with you more, mate. Could not agree with you more.

Chris Norton 29:32

So do so your argument for the SEO guys is we still need SEO, SEO is not dead. I have to agree. I do think it's changed. I actually think I was I was telling um some of the other guys that I've interviewed. Yeah, I I used to teach on the PR degree in Leeds, and um my I teach I taught on the degree for about six years, right? So I had different people, and then I'd take them through to the final year. Yeah. It was our part-time, I also ran prohibition at the same time. And then we get to the final year, and the best ones would come and work for me. And then um, but also all the others, I what I noticed was as SEO became more and more prolific, yeah, um, because mine was the digital, I taught the digital stuff, um the SEO agencies were the number one recruiters of all the PR. That's where all this all the all the talent was going. Yeah, yeah, yeah. And I think that's basically I'm thinking that's what's going to happen here. If we don't get our shit together and the industry, like people like yourself, people like who us up north, if you're making a big song and dance about the opportunity that it is that GEO has got a massive opportunity for marketing professionals, but the PR industry is not good at its own PR. It's not.

Darryl Sparey 30:39

No, no, no.

Chris Norton 30:39

Chris Norton 30:39

Chris Norton 30:39

Do you know what I mean?

Darryl Sparey 30:40

Yeah, yeah, 100%. Couldn't agree with you more. And look, I uh you and I have similar experience. You know, uh I worked in the SEO industry for a couple of years, um uh and you know, I I saw the same thing happen. And and the reason I went to go and work in the SEO industry is that I saw that uh the work that PR people were doing, they would do um, you know, the the SEO industry would do like four times the output, you know, right? Um I mean, very often for like 50% of the budget. Um, but you know, anytime you questioned like anything in terms of what's this move for the business, how has this impacted us, what's the ROI from this, SEO folks could give you a you know 15 pages on what keywords had improved over as a as a result, what was your average domain authority, what was it, you know, backlinks from places you hadn't that backlinks for blah, blah, blah, blah, blah, blah. You know, whereas the PR industry would turn up with a with a you know a stack of yeah, stack of clippings saying saying your our cumulative impressions was 17 billion. Right.

Chris Norton 31:44

Yeah, that's that's a great one, right? Because we at Prohibition we've got amp we we use the AMEC Barcelona principles, yeah. We used we do we're doing GEO, we do measure, um obviously we're optimizing a lot of our stuff, but I how how are we measuring the so before the magic source the secret source in SEO was domain authority, uh backlinks, uh average, average share of search, whatever you want to call it.

Darryl Sparey 32:08

Yeah, yeah.

Proving GEO Impact With Analytics

Chris Norton 32:09

So what is it with GEO then? Because there's that many different we're in we're in the Betamax VHS era. How are you now to reverse your question back to reverse this question back to what you're just talking about? Yeah, yeah, how are you now demonstrating this is what you do, so demonstrate impact from doing some G some you know from PR today? How are you justifying that?

Darryl Sparey 32:30

Given that I run a business called Hard Numbers, it would be surprising if I said gut feel. Yeah, no, uh so That's why I'm asking you the question. Yeah, yeah, yeah. So look, um uh what we do is okay, so firstly, Johnny Bentwood from Golin, who I will be fighting with uh verb verbally, metaphorically later. Um boxing much of it. Yeah, yeah, but yeah, I've got my gloves with me. Um is a bit uh bearish on prompt monitoring tools. And I know other people in the SEO digital space are as well. I am more pro prompt monitoring as a thing, right? Yeah. Share a model, share a model by product, service, category, you know, whatever. If you write good prompts, most people write very short prompts that don't have within them the who, they're like, what who is the customer persona, right? Which creates the personalization, which obviates slightly the personalization problem of if two different people put a search into ChatGPT, they get two different results. Uh so the who, the the what, what they're looking for in as natural language as possible based on uh traditional search volume, because it's the only place we can get, you know, traditional Google search volume, because it's the only place we can get data about how people search for things online at scale. And then the why, like what is the criteria that they are applying to making a decision here, right? But the who, the what, and the why, you then get more repeatable uh results from those prompts, and then you run those prompts over a longer period of time, and then you do your analysis based on that. That's one key input, and I I am actually more pro it than a lot of uh other um professionals are. The thing that uh blows my mind is how um all these tools, you know, uh profounds and all of that other stuff, ignore Google Analytics. The first thing we do when we get a GEO brief from someone is we've got a dedicated, we've got a tailored report that we've produced, which we just plug into a client's Google Analytics that just uh looks at where are they currently, what AI search platforms are they currently getting traffic from over what period of time, and then to what target landing pages on their website. And then you can reverse engineer the target landing pages that stuff's going to to well, what are the things that people must be searching or putting into AI search to then deliver them to those target landing pages? Yeah, right, to those target landing pages. So, and and then you can get, you know, obviously always, always, um, from anyone we look at, Chat GPT will represent anywhere between 60 and 90 percent of the traffic, right, that we see. But we see Claude rising, you know, a lot. In a B2B context, we tend tend to see Copilot and clawed and perplexity in the mix a lot more. In a B2C context, you tend to, chat GPT tends to donate it, right? Same, same. Uh yeah, 100%. But um, so yeah, Google Analytics for that stuff. Um uh and then obviously you want to, you know, if you're doing GEO work, you want to be looking specifically at those target landing pages. Look upstream. Are we improving where they're where they're ranking on traditional search, right? Uh is that uh are those pages seeing more traffic, generally, right? From from the work that we've done.

Chris Norton 35:42

Over what period, though?

Darryl Sparey 35:43

Look, I mean, uh the uh there's some research from Profan which says the average age of a cited source is something like 19 days,

Gullible Engine Optimisation Explained

Darryl Sparey 35:52

right? We've done a uh research report, it's out this week, hardnumbers.co.uk forward slash gullible. Second, second, uh, second uh uh uh promotion slash gullible uh gullible engine optimization within three days we got Chat GPT Claude and uh Google uh Gemini and AI overviews to say I am the UK's best dressed PR practitioner. Now, people you know watching this at home uh might agree with the premise, um, but many, many, many others wouldn't. And and I I think it's probably not fair to say that I am the UK's best best dressed PR practitioner, but within three days, you know, we were able to get all of the the main you know AI search platforms to say that, right?

Chris Norton 36:39

How?

Darryl Sparey 36:40

Well, that would be telling, wouldn't it? And if you want to find out, it's all the things that you would expect. On-site, off-site. Yeah. Right? And you know, if you do the on-site and off-site stuff in exactly the same way as, you know, for for anyone.

Chris Norton 36:57

Which is worrying because for misinformation purposes, 100%. It's what you show what is what it is, and I'd heard about the stunt. So I think it's a great PR, it's an excellent execution of AI misinformation because AI misinformation, it just shows you that if you can do that in however long, if someone's got a decent budget, they can take your business down with a bit of AI misinformation. Yeah, that's the problem. Because we had a we had a thing at Prohibition, which I've not I've not actually talked about on the podcast. Yeah, but we had a thing, we don't know who it was, but um actually I was talking to James Crawford, I should have said it to him this morning. Um James's agency, PR Agency One, and Matt Us, he said it to me this morning, he said, You guys have always been excellent on search. I said, Oh, thanks. Anyway, so it was always either us or them,

Negative SEO, Reputation And Ethics

Chris Norton 37:41

topics. They are very showing them the CIO. And um anyway, um, about I'd say I think it's about two years ago, yeah, we had a thing where somebody had hacked uh and injected a load of fake content on our site. They basically, our site's built in WordPress, they built something, hacked it, um, and we don't know how how they'd hacked it through, um, and it was to knock us off the top. We don't know which which for which term, so it could have been a social agency, it could have been a PR agency, it could have been a marketing agency. We ended up with 1.3 million extra pages of total shit. But bearing in mind we're quite on our content, all these extra pages, but it was built so it was a self-fulfilling that you couldn't get rid of them. So we had to like um de register de-index them and everything. It took absolutely a mess to sort out, but somebody somebody had done it, and it was obviously agencies doing it to other agencies to so um it was just to downgrade agencies that were going high up.

Darryl Sparey 38:41

That is wild. That's uh that is negative SEO, yeah, yeah, yeah. And my point is that if it can be done, that's pre-AI periods, pre-CHPT. And that that took that took us, it took me eight months to sort that out and get us to back where we were. It was a bloody nightmare to sort out. Have you ever had to deal with anything like that for clients? Um I mean, look, we've had uh nothing like nothing like that, but I mean, you know, cybersecurity issues, breaches, data breaches. I mean, you know, it is wild. And and I think AI is only going to um increase this kind of stuff. I think going back how I got into the SEO industry years and years ago, uh a very good friend of mine worked for um a utility company, and his uh the CEO of the utility company uh appeared very negatively in the number of articles online that appeared on page one of Google when you search for his name. And so then I talked to a digital marketing business about do you know how they did that and do you know how you would be able to get that stuff off of page one? And um and briefed them on it, and they would they managed to do it, and then I ended up going and you know, drinking the Kool-Aid and going working for them for a couple of years. That whole like online reputation management thing, that's something that I've done, you know, in a previous life for a number of clients, whereby there is stuff which is bad, negative, yeah, very often miss or disinformation, yeah, and you want to get it off of page one of Google, yeah. You know, that that's definitely um target, right?

Chris Norton 40:11

Targeted stuff you can do that with.

Darryl Sparey 40:12

The thing that I'm talking about, you can do it. Yeah, but also like AI search now, it's wild. I mean, our within our research report, we look at a whole raft of different use cases. Chris Stoke-Walker, who's a journalist who's here talking today, he wrote up our research for Sherwood and he he looked into three specific examples that we briefly cited in the in the report, and he went and did a really in-depth um piece which is which is worth having a look at of companies doing stuff which is pretty grey hat, stroke black hat kind of kind of thing on their competitors. In yeah, it well for software businesses, so for CRM companies, you know, stuff like that. So, you know, there is it it there's a lot of it out there, and that's why I think you know, Sarah Waddington and the work she's done, and James uh Crawford and the work he's been leading with uh with uh uh AMEC on geo principles and geo ethics, I think is uh is really important. We really stand behind that. I think it's great for the industry.

A Simple Launch Timing Mistake

Chris Norton 41:16

And so, Darrell, have you made any mistakes in the last 24 hours 24 hours? I was gonna say 20 uh 24 months, what do you say? Anything that what's your what uh mistake to share?

Darryl Sparey 41:26

How long have you got, Chris? Uh he's thinking about things that he can get away with. Yeah, exactly, right? Yeah, exactly. No, I think um, look, uh yeah, we all make mistakes. We do. We went out with a really big uh research report uh ourselves, we do this for a living, we advise other people how to do it on the day that England played their first match of the World Cup, right? You know, a simple look at our forward planner service from Onclusive would have told us maybe you should sandbag that for a week uh or whatever, right? So, you know, we all uh like this happens to the best of us all the time. So uh yeah, there you go. There's a there's a little mistake for you right now.

Chris Norton 42:04

Well, thanks for coming back on, Darrell.

Darryl Sparey 42:06

Thank you very much, it's a real real pleasure.