Latest / The STANDOUT Agency Podcast / EPISODE 8: Where NOT to use AI in your agency.
Transcript
- 0:04SPEAKER_00: Welcome to The STANDOUT Agency Podcast, for agency owners who want to grow, not just survive.
- 0:09SPEAKER_01: I'm Gareth Healey.
- 0:10SPEAKER_01: I'm an ex-agency owner and author of STANDOUT OR DIE.
- 0:14SPEAKER_01: I help agency founders see where they're still a bottleneck in their own business
- 0:17SPEAKER_01: and make sure they do something about it.
- 0:19SPEAKER_00: And I'm Callum Healey.
- 0:20SPEAKER_00: I find where agencies are losing hours to manual work and then build AI systems
- 0:24SPEAKER_00: that get the time back.
- 0:27SPEAKER_01: Each week, one of us interviews the other on a subject related to agency growth.
- 0:31SPEAKER_01: Everything we get into comes straight out of our own experience and client work.
- 0:34SPEAKER_01: Occasionally we bring in some guests to add their expert perspective on a topic.
- 0:41SPEAKER_00: Welcome back to The STANDOUT Agency Podcast.
- 0:44SPEAKER_00: This is episode eight, where not to use AI in your agency.
- 0:49SPEAKER_00: I think it's important in any discipline to think about what you shouldn't do just
- 0:54SPEAKER_00: as much as what you should do.
- 0:57SPEAKER_00: The essence of strategy.
- 0:59SPEAKER_00: Absolutely.
- 1:00SPEAKER_00: And in this episode, for our listeners, expect to learn and for us to discuss...
- 1:06SPEAKER_00: The fear around AI replacing jobs in the future,
- 1:10SPEAKER_00: the apprenticeship paradox and how judgment is built up over time in careers,
- 1:15SPEAKER_00: why AI can't be trusted to make the final decision,
- 1:19SPEAKER_00: and just because it can ingest every framework known to man,
- 1:23SPEAKER_00: it can't be in the room with you and have that experience firsthand.
- 1:29SPEAKER_00: Exciting.
- 1:30SPEAKER_01: Shall we jump in?
- 1:31SPEAKER_01: Let's do it.
- 1:33SPEAKER_01: To go to the first point,
- 1:35SPEAKER_01: the fear of AI replacing jobs,
- 1:38SPEAKER_01: we're recording this in August 2026,
- 1:40SPEAKER_01: and a couple of the big networks have just reported their recent set of results,
- 1:46SPEAKER_01: WPP,
- 1:46SPEAKER_01: S4 Capital,
- 1:48SPEAKER_01: and as part of that,
- 1:49SPEAKER_01: Or prior to that, both of them had downscaled their staff numbers.
- 1:56SPEAKER_01: They didn't put either of them down totally to AI.
- 1:58SPEAKER_01: You wouldn't be explicit about that kind of thing, of course.
- 2:02SPEAKER_01: Of course, but there may be some aspects of that.
- 2:04SPEAKER_01: But the reality is there is a lot of fear around AI replacing jobs.
- 2:08SPEAKER_01: That might be a sign to some degree that it is.
- 2:12SPEAKER_01: But if there's fear around AI replacing jobs,
- 2:15SPEAKER_01: Callum, you're working with agencies to deploy AI, are you helping them to do that?
- 2:22SPEAKER_01: Are you part of the problem or part of the solution?
- 2:24SPEAKER_00: I hope not.
- 2:26SPEAKER_00: No, that's not my clear objective.
- 2:28SPEAKER_00: That's not my first thought when I help agencies.
- 2:31SPEAKER_00: And just as much as I don't want it taking other people's jobs,
- 2:34SPEAKER_00: I don't want it to be taking mine.
- 2:37SPEAKER_00: And I think that when you...
- 2:40SPEAKER_00: Think about business margins.
- 2:42SPEAKER_00: If you reduce headcount and increase output, there's profit.
- 2:45SPEAKER_00: Brilliant.
- 2:46SPEAKER_00: But I think you can do this without decreasing the headcount.
- 2:49SPEAKER_00: And it's all about making processes efficient,
- 2:52SPEAKER_00: not replacing people, just giving them a tool to do processes better.
- 2:57SPEAKER_00: I get blown away by AI's utility the more I use it and the more I teach people how to use it.
- 3:03SPEAKER_00: But I also have less trust in it the more I learn.
- 3:09SPEAKER_01: Why do you trust it less the more you learn?
- 3:11SPEAKER_01: Surely you trust it more, do you?
- 3:13SPEAKER_00: Yeah, so ultimately it's hallucinations which are a big part of these models and
- 3:22SPEAKER_00: while they appear coherent,
- 3:25SPEAKER_00: they're programmed to and How these models work, simply,
- 3:31SPEAKER_00: we've discussed this in episodes beforehand,
- 3:34SPEAKER_00: they're completely based on probability,
- 3:37SPEAKER_00: and that probability cannot be a certainty in any case.
- 3:41SPEAKER_00: It is the model's best guess.
- 3:44SPEAKER_00: Of what goes on rather than fact.
- 3:46SPEAKER_00: Cannot verify that information without someone being there to verify it for them,
- 3:51SPEAKER_00: which is why someone using the tool is so important for that verification step.
- 3:57SPEAKER_00: It's the layer on top.
- 3:59SPEAKER_00: Like in journalism,
- 4:00SPEAKER_00: the writer or the first drafter of a journal article,
- 4:05SPEAKER_00: the editor would then come on and check the output quality of that thing
- 4:09SPEAKER_00: and I think that's where humans will sit on top of AI's initial drafts.
- 4:14SPEAKER_00: So the human in the loop is,
- 4:16SPEAKER_00: in your opinion,
- 4:17SPEAKER_00: important now and is
- 4:18SPEAKER_01: going to continue to be important going forward?
- 4:20SPEAKER_00: Yeah, exactly.
- 4:22SPEAKER_00: As I said and touched on briefly there,
- 4:24SPEAKER_00: these models aren't built to say I don't know and get it wrong.
- 4:30SPEAKER_00: Companies almost don't want their models to admit that and it makes them look bad,
- 4:34SPEAKER_00: like the technology isn't good enough.
- 4:36SPEAKER_00: They are programmed to guess and give their best attempt at what
- 4:39SPEAKER_00: the right answer is,
- 4:40SPEAKER_00: which is why it can fabricate lies and appear very confident in its findings.
- 4:47SPEAKER_00: I think it's analogous to a multiple choice quiz,
- 4:50SPEAKER_00: where there's four answers usually,
- 4:53SPEAKER_00: it's better to guess if you don't know the answer.
- 4:56SPEAKER_00: That's what AI does, it gives its best shot.
- 4:59SPEAKER_01: I've often thought recently that when people say AI hallucinates,
- 5:03SPEAKER_01: that's a common
- 5:04SPEAKER_00: It's a term and thing now,
- 5:06SPEAKER_01: isn't it?
- 5:07SPEAKER_01: But actually, to my mind, it doesn't really hallucinate at all because it can't think.
- 5:12SPEAKER_01: Yeah, there you go.
- 5:14SPEAKER_01: It actually doesn't hallucinate, it just makes mistakes.
- 5:17SPEAKER_00: It makes the wrong choice.
- 5:20SPEAKER_00: It's the way that these models will always be unless the underlying architecture changes.
- 5:27SPEAKER_00: So Transformers,
- 5:30SPEAKER_00: the recent architecture in 2017 or 2018,
- 5:33SPEAKER_00: a bunch of Google researchers got together and made this revelation.
- 5:37SPEAKER_00: That's where we've seen all these LLM chatbots that are so good at interacting
- 5:42SPEAKER_00: and ingesting information and so on.
- 5:45SPEAKER_00: But the underlying architecture is based on probability.
- 5:48SPEAKER_00: And so it will just guess.
- 5:50SPEAKER_00: And a lot of the time it's correct.
- 5:52SPEAKER_00: Most of the time it is.
- 5:53SPEAKER_00: Sometimes it can slip through the gaps and you're not quite sure.
- 5:57SPEAKER_00: I went to an AI networking event when one of the people I was speaking
- 6:01SPEAKER_00: to was talking about exactly this.
- 6:03SPEAKER_00: And in the context of agencies or any business,
- 6:06SPEAKER_00: if there isn't 100% reliability in something,
- 6:09SPEAKER_00: it can't be trusted to work on its own.
- 6:11SPEAKER_00: Because if it processes important information and you just let it go off
- 6:16SPEAKER_00: and do what it wants, you could get into big trouble.
- 6:19SPEAKER_00: That's why these things like agentic softwares like OpenClaw,
- 6:24SPEAKER_00: you hear softwares like this where people buying Mac Minis,
- 6:27SPEAKER_00: It's really hard to let AI just have access to everything and go, off you go, right.
- 6:31SPEAKER_00: I know you make mistakes sometimes, but do your best.
- 6:35SPEAKER_00: And he was saying that it was an analogy of credit cards and the technology of
- 6:39SPEAKER_00: the microchips within them.
- 6:41SPEAKER_00: It's been the same,
- 6:42SPEAKER_00: or I think they're making a new change to them recently,
- 6:46SPEAKER_00: but for the last 40 years, they'd used the same technology.
- 6:48SPEAKER_00: Very basic technology, but it worked.
- 6:51SPEAKER_00: And 100% reliability is what we want.
- 6:54SPEAKER_01: Increasingly,
- 6:55SPEAKER_01: us humans need or are relying on the machines,
- 7:00SPEAKER_01: but we believe that machines still need us now for the foreseeable.
- 7:06SPEAKER_01: The human in the loop jobs are not going to take over.
- 7:10SPEAKER_00: Yes.
- 7:10SPEAKER_00: Just yet.
- 7:11SPEAKER_00: Exactly.
- 7:12SPEAKER_00: With the way that they are right now, I would say it's impossible for them to replace jobs.
- 7:18SPEAKER_00: And there are some,
- 7:19SPEAKER_00: so there are caveats in radiology,
- 7:22SPEAKER_00: for example,
- 7:23SPEAKER_00: that when the human naked eye of looking at It's better at some things,
- 7:29SPEAKER_00: but we still need to be the ones in control,
- 7:32SPEAKER_00: using it for those things that it's so good at.
- 7:58SPEAKER_01: Callum, if we're saying that clearly AI has got massive potential here,
- 8:02SPEAKER_01: is already fulfilling some of it,
- 8:03SPEAKER_01: and it's going to be an even more fundamental part of people's jobs,
- 8:06SPEAKER_01: but not replace, certainly not all of them.
- 8:09SPEAKER_01: We were talking there about how judgment of the human,
- 8:13SPEAKER_01: whether things are right or wrong,
- 8:15SPEAKER_01: whether things are good or bad,
- 8:16SPEAKER_01: whether the information the AI is producing is useful or not,
- 8:21SPEAKER_01: is still really important, hugely important.
- 8:24SPEAKER_01: Verification step, of course.
- 8:26SPEAKER_01: But a lot of that,
- 8:27SPEAKER_01: unless it's just saying yes or no and a basic judgment call,
- 8:33SPEAKER_01: the AI can probably do that.
- 8:34SPEAKER_01: But if it's a more considered judgment call,
- 8:37SPEAKER_01: it strikes me that the people using AI now,
- 8:41SPEAKER_01: maybe talking earlier about hospitals using AI scans,
- 8:47SPEAKER_01: the human in that situation now will be using AI, but they'll be a very experienced
- 8:56SPEAKER_01: Medical professional who have done it themselves without AI for many years
- 8:59SPEAKER_01: and therefore can apply their judgment to it.
- 9:01SPEAKER_01: But what happens do you think when in some years,
- 9:06SPEAKER_01: fast forward in a little bit,
- 9:08SPEAKER_01: but not too soon,
- 9:09SPEAKER_01: when the newer generations of the workforce come along and start to use AI natively,
- 9:15SPEAKER_01: of course,
- 9:16SPEAKER_01: but also don't accumulate those That expertise,
- 9:20SPEAKER_01: that experience, don't,
- 9:22SPEAKER_01: you know,
- 9:23SPEAKER_01: we often mention it,
- 9:23SPEAKER_01: but we're gym goers,
- 9:25SPEAKER_01: don't put in the reps,
- 9:26SPEAKER_01: don't put in the exercises, just come straight in and start using AI.
- 9:30SPEAKER_01: That judgment is not going to be credible, is it?
- 9:33SPEAKER_00: Yeah, it's almost like giving someone the best shape of their lives immediately and
- 9:38SPEAKER_00: then going, good luck maintaining it, because they don't know how to get that.
- 9:43SPEAKER_00: The expertise is so important, it's about...
- 9:46SPEAKER_00: Accumulating the knowledge of what good output looks like in a specialised field.
- 9:52SPEAKER_00: Because of the nature of engine room work that you'd put juniors into first,
- 9:57SPEAKER_00: it's the ones that have the least amount of error, I would say, in most fields.
- 10:03SPEAKER_00: It's usually quite laborious and manual,
- 10:06SPEAKER_00: perhaps,
- 10:07SPEAKER_00: work that senior team members don't want to be doing,
- 10:10SPEAKER_00: because they've had their time doing it before, they've been in the coalface.
- 10:13SPEAKER_00: True, but it's a learning experience going through, or it has been for years.
- 10:18SPEAKER_01: Hundreds of years, to agree, for humans to do that, to gain their stripes on a certain topic.
- 10:24SPEAKER_00: Exactly,
- 10:24SPEAKER_00: so you need to go through that period in order to build that judgment,
- 10:29SPEAKER_00: to be able to verify what AI's output is, if it's good or if it's bad.
- 10:35SPEAKER_00: There's no possible way of doing this if you haven't accumulated that knowledge.
- 10:39SPEAKER_00: And it's why agencies hand...
- 10:44SPEAKER_00: AI the keys to that workflow because it is the one that's the most simple.
- 10:50SPEAKER_00: I guess this is all about change management ultimately and as time has gone
- 10:54SPEAKER_00: on there's been multiple shifts in society and one of those was sort of
- 10:59SPEAKER_00: the digital revolution back in the 2000s when Macbooks and email
- 11:04SPEAKER_00: and all this was coming out and you live through that in your own agency and
- 11:09SPEAKER_00: in my eyes Digital or computers didn't replace anyone.
- 11:13SPEAKER_00: Maybe that was the discussion at the time.
- 11:15SPEAKER_00: It seems silly that computers would replace people now.
- 11:18SPEAKER_00: So maybe a little bit about your story and a couple of decades ago of that change.
- 11:23SPEAKER_01: Yeah, a couple of decades ago.
- 11:25SPEAKER_00: Why would you put it like
- 11:26SPEAKER_01: that?
- 11:27SPEAKER_01: But of course, during my career, I have lived through that.
- 11:30SPEAKER_01: Despite what I might look like,
- 11:32SPEAKER_01: I didn't actually,
- 11:33SPEAKER_01: when I first started in agencies,
- 11:34SPEAKER_01: which was after university, it was just after My first agency was digitized.
- 11:42SPEAKER_01: It was a Mac studio.
- 11:43SPEAKER_01: It wasn't the drawing boards and the artist drawing boards that people often talk
- 11:48SPEAKER_01: about and some people still experience.
- 11:49SPEAKER_01: I was a little bit after that, but it was at the cusp of that.
- 11:53SPEAKER_01: It was very early days in that sense.
- 11:55SPEAKER_01: So definitely witnessed that in terms of artwork,
- 11:58SPEAKER_01: studios becoming digitized through Macs and of course email and all that sort
- 12:02SPEAKER_01: of stuff, digital communication and then digital marketing, digital marketing.
- 12:07SPEAKER_01: Marketing tools, of course, came on board.
- 12:09SPEAKER_01: So it lived through all that.
- 12:11SPEAKER_01: Did it replace jobs?
- 12:14SPEAKER_01: I don't recall it did.
- 12:15SPEAKER_01: It refocused jobs.
- 12:17SPEAKER_01: People had to,
- 12:19SPEAKER_01: if not retrain,
- 12:19SPEAKER_01: they had to sort of refocus on what they were doing and started to do their work
- 12:24SPEAKER_01: in different ways, using different new tools.
- 12:27SPEAKER_01: It strikes me that we went from,
- 12:30SPEAKER_01: in that period,
- 12:31SPEAKER_01: a very literal hands-on,
- 12:34SPEAKER_01: going back to the drawing boards and artwork,
- 12:36SPEAKER_01: somebody actually moving pens around and cutting out letters to somebody using
- 12:41SPEAKER_01: a mouse and a keyboard.
- 12:44SPEAKER_01: For many people,
- 12:44SPEAKER_01: that was a big change,
- 12:45SPEAKER_01: but it just created the ability to do more work,
- 12:48SPEAKER_01: more accurate work, cut out a lot of the steps.
- 12:50SPEAKER_01: I think AI will do the same.
- 12:53SPEAKER_01: I think the interface is interesting because I think,
- 12:55SPEAKER_01: as we're learning,
- 12:56SPEAKER_01: the interface might move from literal hand-operated
- 13:02SPEAKER_00: inputs to keyboard inputs
- 13:05SPEAKER_00: to maybe voice inputs.
- 13:06SPEAKER_00: To voice, yeah, that's a big thing.
- 13:08SPEAKER_00: We've learned that using Whisperflow, which we love, so a little shout out to them, of course.
- 13:13SPEAKER_00: You can just prompt quicker when speaking and it allows you to articulate your thoughts more.
- 13:18SPEAKER_00: So, yeah,
- 13:18SPEAKER_00: as time goes on,
- 13:20SPEAKER_00: We people in the agency world, in this case, figure out better ways of doing things.
- 13:25SPEAKER_01: If Whisperflow are out there, by the way, we'd be happy for them to sponsor this podcast.
- 13:29SPEAKER_00: Oh, of course.
- 13:30SPEAKER_00: Yes, please.
- 13:30SPEAKER_00: That'd be great.
- 13:31SPEAKER_00: I've signed up to a year already.
- 13:33SPEAKER_00: I've committed myself.
- 13:35SPEAKER_00: It's interesting because the digital revolution,
- 13:38SPEAKER_00: there was a mouse and keyboard, sure, but it wasn't something that had intelligence.
- 13:43SPEAKER_00: AI is artificial intelligence,
- 13:45SPEAKER_00: so there's no wonder that people are feeling slightly concerned about
- 13:50SPEAKER_00: the We've never had on earth something that is more intelligent than us,
- 13:58SPEAKER_00: even if it is just at certain things.
- 14:00SPEAKER_00: Luckily,
- 14:00SPEAKER_00: it gets its intelligence from the things that we've created and put online and
- 14:04SPEAKER_00: the data we feed it.
- 14:06SPEAKER_00: It can just process it very quickly.
- 14:08SPEAKER_00: I think the crucial lesson here Rather than just having AI and agents running
- 14:14SPEAKER_00: the cold face,
- 14:15SPEAKER_00: which as we've covered would be a nightmare anyway
- 14:17SPEAKER_00: because they'd start getting things wrong that compile that over multiple years
- 14:21SPEAKER_00: and you've got yourself or even weeks,
- 14:24SPEAKER_00: little problems here and there can And make workflows completely break down.
- 14:28SPEAKER_00: I would say instead of learning from AI and getting AI
- 14:31SPEAKER_00: to tell you how things should be done,
- 14:33SPEAKER_00: which is just the default average of all the data it's trained on,
- 14:37SPEAKER_00: you should be learning through AI.
- 14:40SPEAKER_00: In this case, we've talked a lot about context, of course, in past episodes.
- 14:44SPEAKER_00: The senior leaders who have the judgment now and have built up those reps need to be
- 14:49SPEAKER_00: the ones configuring the systems that contains the schema
- 14:52SPEAKER_00: of what they've learned throughout their career.
- 14:55SPEAKER_00: And so when AI and this AI system that you've created,
- 14:59SPEAKER_00: and I say system,
- 15:00SPEAKER_00: it could be a custom GPT or a clawed skill,
- 15:04SPEAKER_00: but you tell it the difference between right and wrong,
- 15:06SPEAKER_00: what good output looks like,
- 15:08SPEAKER_00: what bad output looks like, what to say, what not to say.
- 15:12SPEAKER_00: Of course, it's very...
- 15:14SPEAKER_00: Related to your workflow.
- 15:16SPEAKER_00: Every workflow is different.
- 15:17SPEAKER_00: There's practically billions out there that's possible.
- 15:20SPEAKER_01: But it is related to that.
- 15:22SPEAKER_01: We've got experience.
- 15:22SPEAKER_01: You've got experience live this week.
- 15:26SPEAKER_01: As bizarre as this might sound, you've got to understand your workflow, haven't you?
- 15:30SPEAKER_01: Of course.
- 15:30SPEAKER_01: That's the big shift.
- 15:31SPEAKER_01: You've got to be able to describe it, communicate it.
- 15:33SPEAKER_01: You've got to be able to draw it out.
- 15:36SPEAKER_01: And a lot of the agencies that we're talking to, and I'm sure many others,
- 15:45SPEAKER_01: The workflow that they have happens.
- 15:48SPEAKER_01: They do great work.
- 15:48SPEAKER_01: The team produces great work.
- 15:53SPEAKER_01: But it's natural.
- 15:55SPEAKER_01: People don't often have the workflows nailed down and documented.
- 16:03SPEAKER_01: It's the human team working together and things just happen and people know what
- 16:09SPEAKER_01: to do, but almost naturally.
- 16:13SPEAKER_01: We've talked about this a lot this week.
- 16:14SPEAKER_01: If you're going to get AI involved, you've got to go a step beyond that.
- 16:20SPEAKER_01: Scope out your processes and workflows.
- 16:22SPEAKER_00: In painful detail, may I add.
- 16:24SPEAKER_00: And that's one of the biggest shifts, I would say.
- 16:27SPEAKER_00: Being new to the agency world,
- 16:29SPEAKER_00: I was shocked about how much processes were trapped in people's heads.
- 16:33SPEAKER_00: And that's not to say it's a bad thing.
- 16:35SPEAKER_00: I think a lot of agencies are successful because of that
- 16:39SPEAKER_00: and they trust their team members who have the knowledge.
- 16:41SPEAKER_00: But with AI as the one that's helping us speed up processes along the way,
- 16:48SPEAKER_00: we need to make sure AI is very clear on what the process is.
- 16:51SPEAKER_00: And I think that's a great topic for a future episode.
- 16:54SPEAKER_00: We could talk a lot about that and the shift in landscape there.
- 16:58SPEAKER_01: We've often reflected in other industries,
- 17:01SPEAKER_01: non-agency industries,
- 17:02SPEAKER_01: in pharmaceuticals, for example, those processes have to be clear.
- 17:08SPEAKER_01: You don't produce a drug without being very clear.
- 17:13SPEAKER_01: Agencies work with farmers,
- 17:14SPEAKER_01: farmer business of course,
- 17:15SPEAKER_01: but it's just one example,
- 17:16SPEAKER_01: but it isn't that rigid and I think that's fuelled and perpetuated things
- 17:19SPEAKER_01: in people's heads really.
- 17:38SPEAKER_00: Yeah and it is an interesting topic although we will make a future episode about it.
- 17:44SPEAKER_00: I think it stems from risk.
- 17:46SPEAKER_00: You've talked to me a lot before about how in the agency world in general,
- 17:50SPEAKER_00: there are definitely a lot of things that can go wrong.
- 17:53SPEAKER_00: Insurance claims,
- 17:54SPEAKER_00: there's law breaching contracts and losing clients,
- 17:56SPEAKER_00: but in general,
- 17:58SPEAKER_00: no one's going to die as a result or no one's going to get injured as a result.
- 18:03SPEAKER_00: Whereas in pharmaceutical companies,
- 18:05SPEAKER_00: they have to make the drug exactly how they've designed and tested it so
- 18:08SPEAKER_00: that nothing No one gets
- 18:11SPEAKER_01: injured or hurt.
- 18:11SPEAKER_01: Lives are on the line, quite literally, aren't they?
- 18:14SPEAKER_00: People's brains are on the line now, with AI coming in.
- 18:17SPEAKER_00: That's what we're kind of talking about here.
- 18:19SPEAKER_00: It's important that you use AI to amplify your thinking rather than replacing it,
- 18:25SPEAKER_00: because in 20 years we don't want to be having a society filled with drones
- 18:30SPEAKER_00: that can't think without having an AI chatbot next to them.
- 18:34SPEAKER_01: So we've talked a little bit about judgment and the need for judgment using AI,
- 18:39SPEAKER_01: but there's also, in my mind, there's sort of decision making itself, isn't there?
- 18:46SPEAKER_01: That decision-making in many agencies,
- 18:49SPEAKER_01: as we are talking about,
- 18:53SPEAKER_01: often is based on facts,
- 18:54SPEAKER_01: but it is also the nuance of human interaction and
- 18:58SPEAKER_01: whether those decisions are appropriate for the right stage that people are in,
- 19:03SPEAKER_01: whether they are the right decision at the right time.
- 19:06SPEAKER_01: And AI, because it is not human, Can't make that sort of judgment itself, can it?
- 19:13SPEAKER_01: It will fail to take into account, for example, the stresses and strains of running an agency.
- 19:20SPEAKER_01: Do we really need to face into this particular decision now when we're very focused
- 19:25SPEAKER_01: on trying to solve this problem over here or get this piece of client work out
- 19:28SPEAKER_01: the door for five o'clock?
- 19:31SPEAKER_00: Yeah, it's strange really because knowledge itself and words, we created them.
- 19:36SPEAKER_00: And humans,
- 19:37SPEAKER_00: it's the idea that humans are only ever going to interact with humans in business
- 19:41SPEAKER_00: as far as we're aware of where reality sits right now.
- 19:44SPEAKER_00: But we created words and AI can understand those words and the facts
- 19:49SPEAKER_00: that are our construct,
- 19:50SPEAKER_00: but it doesn't understand the people and how we interact with each other.
- 19:55SPEAKER_00: So our brains have evolved for sort of 200,000 years.
- 20:02SPEAKER_00: And we've evolved to survive ultimately,
- 20:06SPEAKER_00: but in that process for evolutionary purposes,
- 20:10SPEAKER_00: understanding other humans was essential to that,
- 20:13SPEAKER_00: whether that's through friendships, starting families, avoiding...
- 20:19SPEAKER_00: Enemy tribes and conflicts.
- 20:21SPEAKER_00: So our brains are completely designed to pick up on subconscious human cues.
- 20:27SPEAKER_00: Whether we're aware of it or not,
- 20:30SPEAKER_00: it's what we're so good at and we're built to interact with each other.
- 20:34SPEAKER_00: An interesting side note on that is that online meetings can be very tiring,
- 20:39SPEAKER_00: more tiring than in-person meetings.
- 20:42SPEAKER_00: And the idea of that is because...
- 20:44SPEAKER_00: Our brains are so used to picking up on subconscious cues and they're a lot harder
- 20:49SPEAKER_00: to read for a screen.
- 20:50SPEAKER_00: So when you're with someone,
- 20:51SPEAKER_00: you're reading their body language,
- 20:53SPEAKER_00: their tone of voice, little mini facial expressions.
- 20:56SPEAKER_00: It's better for our brains because subconsciously in the background, we're processing them.
- 21:01SPEAKER_00: AI clearly can't do this.
- 21:03SPEAKER_00: It doesn't understand human emotions.
- 21:05SPEAKER_00: It can read every psychological study on them, but it's never felt fear, anxiety, stress.
- 21:12SPEAKER_00: Even kindness, warmth, love, things like this.
- 21:15SPEAKER_00: And that means it can give you advice and present the facts to you about
- 21:21SPEAKER_00: a situation, but it doesn't understand actually how that advice should be applied.
- 21:28SPEAKER_00: It's our job to take the facts and assess it through our knowledge and our lens
- 21:34SPEAKER_00: because we can understand other people so well.
- 21:37SPEAKER_01: I'm feeling a bit more comfortable now about the fact that AI is useful
- 21:40SPEAKER_01: but won't be taking everybody's jobs overnight.
- 21:43SPEAKER_01: Exactly.
- 21:43SPEAKER_00: Yeah, it's purely a prediction machine.
- 21:47SPEAKER_00: It's how we created it and it uses our knowledge and it's very quick,
- 21:51SPEAKER_00: which is great, but yeah, that's the biggest factor.
- 21:55SPEAKER_00: Going forward, sorry,
- 21:58SPEAKER_00: agencies should lean into these traits and the idea
- 22:01SPEAKER_00: that AI can facilitate certain processes
- 22:04SPEAKER_00: that would actually dampen those naturally human sort
- 22:09SPEAKER_00: of gifted talents like creativity,
- 22:12SPEAKER_00: strategy,
- 22:13SPEAKER_00: judgment, how personable you are, how friendly you are in meetings and whatnot.
- 22:18SPEAKER_00: It allows us to do more of that while it takes over the more kind of analytical,
- 22:21SPEAKER_00: factual based
- 22:24SPEAKER_01: That sort of relationship stuff has always been a core part of
- 22:27SPEAKER_01: a professional services firm like an agency.
- 22:32SPEAKER_01: As much as clients buy the value delivered,
- 22:36SPEAKER_01: part of that value is in how they work with an external Agency or Supplier
- 22:42SPEAKER_01: in this case, so that's hugely important.
- 22:44SPEAKER_01: So agencies are good at that stuff, so we
- 22:46SPEAKER_00: need to lead into it.
- 22:47SPEAKER_00: It's relationships, isn't it?
- 22:48SPEAKER_00: A lot of the client development work that will help agencies,
- 22:52SPEAKER_00: but it's all about retention and checking up with your clients and making sure all
- 22:56SPEAKER_00: the services are good and if they're okay.
- 22:58SPEAKER_01: Understanding their stresses and strains, pushing when you can.
- 23:03SPEAKER_01: Stepping back a little bit when you know that it's not the right time.
- 23:07SPEAKER_01: All that sort of stuff.
- 23:07SPEAKER_00: Good.
- 23:08SPEAKER_01: So Callum,
- 23:10SPEAKER_01: we've talked in the past and maybe touched a little bit on today the fact
- 23:14SPEAKER_01: that we believe AI is the tool, isn't it?
- 23:17SPEAKER_01: Yeah.
- 23:19SPEAKER_01: But the human in the loop,
- 23:21SPEAKER_01: the operator of the tool is still and will remain still
- 23:24SPEAKER_00: hugely important
- 23:25SPEAKER_00: in our view.
- 23:26SPEAKER_00: Everybody will be pleased to know.
- 23:27SPEAKER_00: We've been flattering human egos in this podcast so far.
- 23:31SPEAKER_00: People listening will be glad to know how important everyone still is.
- 23:35SPEAKER_01: But we need to,
- 23:35SPEAKER_01: as I described earlier in the digital revolution that I experienced in my business,
- 23:40SPEAKER_01: and here's yet another one.
- 23:43SPEAKER_01: That has been dropped on us.
- 23:45SPEAKER_01: We need to adapt as humans, we're good at adapting as human beings, so that's good news as well.
- 23:50SPEAKER_01: But we need to think about how we work,
- 23:55SPEAKER_01: but also maybe how we train ourselves,
- 23:57SPEAKER_01: but also particularly,
- 23:59SPEAKER_01: as we touched upon earlier, our junior people as they come through the ranks.
- 24:02SPEAKER_01: I know you've got an interesting parallel or thought on that from education.
- 24:08SPEAKER_00: Yeah, exactly.
- 24:11SPEAKER_00: In a way,
- 24:11SPEAKER_00: juniors coming in are going to naturally be more tech savvy,
- 24:14SPEAKER_00: so it's quite a benefit that they'll kind of be more adept at using AI in this case.
- 24:22SPEAKER_00: I compare AI as a phenomenon to a calculator.
- 24:27SPEAKER_00: It's a tool that processes information,
- 24:30SPEAKER_00: whatever you input, It then gives an output based on your input, that's the quality.
- 24:35SPEAKER_00: So a mathematician still needs to understand mathematics in order to know how to use
- 24:40SPEAKER_00: the calculator, then it gives the output.
- 24:43SPEAKER_01: And a mathematician could do far more complicated calculus using a calculator than I could.
- 24:47SPEAKER_00: Exactly, yeah, because they understand the specialism.
- 24:50SPEAKER_00: They understand the specialism of mathematics as a discipline.
- 24:54SPEAKER_00: Agencies understand their certain niche and what makes a good...
- 25:00SPEAKER_00: And what makes good input actually,
- 25:02SPEAKER_00: because that's the whole seniority piece that we talked about.
- 25:06SPEAKER_00: These people who have built up the judgment,
- 25:08SPEAKER_00: they know what good input looks like to a problem and a process.
- 25:11SPEAKER_01: A brief,
- 25:12SPEAKER_01: as I would always still call it,
- 25:13SPEAKER_01: I would call it a brief, which is actually as a prompt,
- 25:19SPEAKER_00: isn't it?
- 25:19SPEAKER_01: A brief that agencies have used that term for a long,
- 25:23SPEAKER_01: long time, as long as I've been in the industry, is in essence a prompt.
- 25:28SPEAKER_00: Yeah, so we know how to,
- 25:29SPEAKER_00: so agencies in this case and particularly the senior people have the best idea
- 25:34SPEAKER_00: of what a good prompt would look like in their specialism.
- 25:38SPEAKER_00: I remember back to my school days learning maths and there were two papers.
- 25:45SPEAKER_00: It was mental maths and it was maths with a calculator.
- 25:50SPEAKER_00: It was very separated,
- 25:51SPEAKER_00: even in the lessons that we had,
- 25:53SPEAKER_00: it was very clear,
- 25:54SPEAKER_00: you know,
- 25:54SPEAKER_00: we even got told to buy calculators externally,
- 25:58SPEAKER_00: you know,
- 25:58SPEAKER_00: they didn't provide them,
- 25:59SPEAKER_00: we had to buy this crazy calculator with all these different buttons on.
- 26:03SPEAKER_00: We still had to learn how to use the calculator.
- 26:05SPEAKER_00: As much as I myself might have been decent at mental maths,
- 26:10SPEAKER_00: If I'd never encountered a calculator before, I wouldn't know how to use it.
- 26:13SPEAKER_00: What is this thing?
- 26:13SPEAKER_00: It's the same with AI.
- 26:15SPEAKER_00: You might have the best specialism and the best knowledge in the world.
- 26:18SPEAKER_00: You still need to know how to apply it in the technology.
- 26:22SPEAKER_00: So yeah, I would say that that's the distinction.
- 26:25SPEAKER_00: Agency is the best case scenario.
- 26:26SPEAKER_00: It's about learning the tool.
- 26:29SPEAKER_00: Most importantly,
- 26:30SPEAKER_00: learning how to apply your specific knowledge and get the best out of it
- 26:34SPEAKER_00: and get more efficient processes, better quality.
- 26:37SPEAKER_00: You can spend more time on certain Initiatives and that comes with understanding
- 26:44SPEAKER_00: the kind of do's and don'ts and where AI is good and bad.
- 26:49SPEAKER_00: So Dad, you're an author,
- 26:51SPEAKER_00: which is amazing,
- 26:52SPEAKER_00: that's something that throughout your career you were always thinking about writing
- 26:57SPEAKER_00: a book and it's existed now for, what is it, six years?
- 27:02SPEAKER_01: Probably coming up to, yeah, maybe five years.
- 27:05SPEAKER_01: Is this an opportunity to plug the book?
- 27:07SPEAKER_01: I guess so, yes.
- 27:08SPEAKER_01: Stand out or die, the book, about five years, yeah.
- 27:11SPEAKER_00: Yeah and there is a kind of accumulation of the experience
- 27:18SPEAKER_00: that you've learned throughout your career,
- 27:20SPEAKER_00: the lessons you've learned and it gives people the framework and the guide
- 27:25SPEAKER_00: to approach agencies in the way that you've seen success and how you've gone
- 27:31SPEAKER_00: about your career and We have systems that have taken,
- 27:35SPEAKER_00: for example,
- 27:36SPEAKER_00: the newest STANDOUT framework,
- 27:38SPEAKER_00: although it's STANDOUT or DIE,
- 27:40SPEAKER_00: it's not strictly the system we've got today, perhaps an update in the future.
- 27:46SPEAKER_01: That was version 1 in the book,
- 27:47SPEAKER_01: but we're now on at least version 2, but it's the same principle.
- 27:51SPEAKER_00: So we're updating the system as it goes along,
- 27:54SPEAKER_00: as everyone should, I think that's the nature of AI.
- 27:58SPEAKER_00: AI knows certain things that you ingest into it,
- 28:00SPEAKER_00: but you actually need it to keep it up to date.
- 28:03SPEAKER_00: That's another element and why juniors accumulating knowledge is so important for the future.
- 28:10SPEAKER_00: But back to the book.
- 28:11SPEAKER_00: Anyway, I'm intrigued because You've lived that experience and that's your knowledge
- 28:16SPEAKER_00: and AI can recite it back to you and understand the system and I'm sure
- 28:21SPEAKER_00: because it's naturally quite sycophantic AI it's probably praised it very heavily.
- 28:26SPEAKER_00: But how does it make you feel that you're the one who's created the system
- 28:29SPEAKER_00: but AI can recite it but it's never kind of sat in the room or lived the experience
- 28:34SPEAKER_00: that kind of made you write it?
- 28:37SPEAKER_01: No, totally.
- 28:37SPEAKER_01: So, I mean,
- 28:38SPEAKER_01: we have,
- 28:40SPEAKER_01: in our own sort of environment,
- 28:41SPEAKER_01: we've fed in the system, haven't we, of course, into AI and the book, actually.
- 28:48SPEAKER_01: And that's been huge.
- 28:50SPEAKER_01: I can see the benefits of that,
- 28:52SPEAKER_01: personally,
- 28:53SPEAKER_01: because we can interact with the book,
- 28:55SPEAKER_01: almost, or interact with the system through AI, which is great, and build on it.
- 28:59SPEAKER_01: And as you've already said, Callum, improve it.
- 29:03SPEAKER_01: So that's a fantastic thing.
- 29:07SPEAKER_01: How does it make me feel?
- 29:10SPEAKER_01: I'm positive about that.
- 29:12SPEAKER_01: You could see a situation where,
- 29:14SPEAKER_01: actually,
- 29:15SPEAKER_01: if all the experience and knowledge is there now in AI,
- 29:20SPEAKER_01: does it need me, the
- 29:21SPEAKER_00: author, the originator
- 29:23SPEAKER_01: of that?
- 29:24SPEAKER_01: Or does somebody need me, the author?
- 29:25SPEAKER_01: Can you just use AI?
- 29:28SPEAKER_01: It's a fair challenge, but at present, how do I feel about it?
- 29:33SPEAKER_01: I feel great that it's an extension of the book and we can use it internally
- 29:37SPEAKER_01: that way,
- 29:38SPEAKER_01: but I'm still very positive on the fact that,
- 29:41SPEAKER_01: and bullish on the fact that even if we...
- 29:43SPEAKER_01: It allowed people, which we may do in the future, to interact with the system using AI.
- 29:51SPEAKER_01: I'm still very bullish about the fact that they would still need
- 29:54SPEAKER_01: or want somebody like me or you or in a different business,
- 30:01SPEAKER_01: another consultant using their own model.
- 30:04SPEAKER_01: To interpret,
- 30:07SPEAKER_01: to act as the
- 30:11SPEAKER_01: human judgment and human decision-making before the judgment
- 30:17SPEAKER_01: or decision-making is made.
- 30:20SPEAKER_01: There's lots of principles in my book and you can feed them into AI
- 30:24SPEAKER_01: and you can ask it a question and it will give you the answer based on the system.
- 30:28SPEAKER_01: As we've said earlier,
- 30:30SPEAKER_01: that AI doesn't take into account the Currently human condition,
- 30:38SPEAKER_01: if I'm working with agency clients,
- 30:39SPEAKER_01: it doesn't take into account their current state of mind,
- 30:43SPEAKER_01: whether they're very open-minded to pushing the business forward and looking
- 30:48SPEAKER_01: at new strategies or whether they're under incredible pressure from other areas
- 30:54SPEAKER_01: and really need to Weareagentsofchange
- 31:14SPEAKER_00: It lacks nuance.
- 31:16SPEAKER_00: No matter how much data you feed AI,
- 31:19SPEAKER_00: in this case,
- 31:20SPEAKER_00: it only knows about your methodology and the facts, the theory behind everything.
- 31:27SPEAKER_00: I've learned throughout my career and education,
- 31:32SPEAKER_00: the value is execution and it's the decisions that you make based on knowledge.
- 31:38SPEAKER_00: Knowledge is just part of the equation,
- 31:39SPEAKER_00: it's actually the person who The way you act on it is very different
- 31:45SPEAKER_00: in multiple scenarios.
- 31:46SPEAKER_00: So you've had 30 years of living in the agency world and learning lessons
- 31:52SPEAKER_00: that have shaped the way that you think and behave.
- 31:54SPEAKER_00: That is more valuable than a framework that you've created.
- 31:58SPEAKER_00: The framework is just kind of the output of that, but the source is the biggest value.
- 32:04SPEAKER_00: To agency owners out there or anyone in the agency world who has expertise
- 32:10SPEAKER_00: and have built up credibility and skill among their time, that is the value.
- 32:15SPEAKER_00: It's what you've learned and who you are and the way you think and make decisions.
- 32:20SPEAKER_00: AI can know the facts about a situation,
- 32:23SPEAKER_00: but it doesn't know how to apply the facts,
- 32:25SPEAKER_00: especially since we've covered that it doesn't know how to interact with people
- 32:29SPEAKER_00: because it doesn't know how people think.
- 32:32SPEAKER_00: We talked before about About that and evolutionary biology and how it's all
- 32:39SPEAKER_00: about subconscious cues and interacting with people,
- 32:42SPEAKER_00: we have these neurons,
- 32:44SPEAKER_00: these brain cells called mirror neurons,
- 32:47SPEAKER_00: where when I observe something happening to someone else, I feel what they feel.
- 32:53SPEAKER_00: Their mirror neurons light up and mine do and it's the cornerstone of empathy.
- 32:58SPEAKER_00: That's how people learn vicariously,
- 33:00SPEAKER_00: you know,
- 33:00SPEAKER_00: you see someone else getting shouted out at school,
- 33:02SPEAKER_00: you don't want to feel that same kind of shame and embarrassment that they do,
- 33:06SPEAKER_00: it's empathetic.
- 33:07SPEAKER_00: AI doesn't have any understanding of that feeling of emotion,
- 33:12SPEAKER_00: it can only understand the facts of it, and they're two very different things.
- 33:16SPEAKER_00: It's like, if you remember Star Trek, it's like Mr. Spock, isn't it?
- 33:19SPEAKER_00: Yeah.
- 33:19SPEAKER_01: Without that, you know, the emotion.
- 33:21SPEAKER_01: Yeah.
- 33:21SPEAKER_01: It gives you the fact,
- 33:22SPEAKER_01: very clever,
- 33:23SPEAKER_01: but doesn't have the ability to connect or The emotions to navigate their way
- 33:31SPEAKER_01: through situations.
- 33:33SPEAKER_00: And facts, I can't believe I'm saying this because I'm quite an analytical,
- 33:37SPEAKER_00: factual person,
- 33:38SPEAKER_00: but they aren't as important really because if you can't interact with other people,
- 33:45SPEAKER_00: There's no point in fact.
- 33:47SPEAKER_00: It isn't the basis of human interaction.
- 33:50SPEAKER_00: It's just a part of it.
- 33:53SPEAKER_00: And AI can process that part.
- 33:56SPEAKER_00: It's great at that part.
- 33:57SPEAKER_00: You should be the one applying that knowledge.
- 34:01SPEAKER_01: I think that's a great place to finish.
- 34:02SPEAKER_00: I think so too.
- 34:04SPEAKER_00: Good episode.
- 34:05SPEAKER_00: Thanks, Colin.
- 34:05SPEAKER_00: Always a pleasure, Dad.
- 34:08SPEAKER_00: We'll see you in the next episode.
- 34:10SPEAKER_00: Thanks for watching if you got this far.
- 34:15SPEAKER_00: You've been listening to Gareth Healey and Callum Healey.
- 34:17SPEAKER_00: Together, we are Agents of Change, which is also our URL, weareagentsofchange.com.
- 34:23SPEAKER_01: If you want support with growing your agency,
- 34:25SPEAKER_01: including how AI can help you future-proof your business,
- 34:28SPEAKER_01: visit us online to learn more about what we do and find out how to connect with us.