AI Agents and the Future of Marketing

So one of the, some of the interesting stuff though that we've got and we've learned over the last few years as we've worked more and more with AI, Alistair's done a lot of stuff here and he's a very experienced AI developer, we're sort of like, we're both a little bit ADHD I suppose and therefore in the back of it we're very good at jumping across multiple things which helps when you're running armies of agents.

But, and what Andy talked about, which came out a few months ago, was cloud design, so this came out, that release, when cloud released their cloud design, I went, oh my goodness, this is now a big change in that these AI systems can now make stuff that looks good, right? It's got a view in the world of design.

They had to release a new model. It was Opus 4 .7. For those of you who are interested, they actually power -clawed design before that. It really wasn't very good at doing all the design. It was getting better, but that was a real step change.

And at that point in time, I have also a health AI startup that I run, and we were marketing that and bringing it to market. And what I really wanted to do was scale my marketing efforts around that stuff. So with these things, we suddenly realized that what I really wanted to do was build a harness for my marketing efforts, and that's what we set out to do.

AI, Jobs, and Human Roles

So what has now existed is we're going to talk a little bit more about AI. A lot of stuff in the news and AI over the last week also has been around, it's going to destroy the world.

My 89 -year -old father called me up on Monday morning and said, son, what's this AI stuff you're doing? Are you going to kill us all? That was the question I got.

So how many of you guys are worried about AI at the moment in terms of existentially or even just in terms of your job and stuff? How many has got a bit of fear? Yeah, fair enough. I think that's fine.

Well, I've got some good news to start with. So AI isn't necessarily removing all the humans. Where did I get this from?

So just last week, there was a big report came out in Silicon Valley by a company called Box. 1600 large companies but they talked about who's using the AI well, who's using AI not so well and what do those look like. And the good news is the companies using the AI really well were actually adding jobs. They weren't using AI to cut loads of jobs, they were actually using AI to add jobs.

And what was the nature of those jobs that were being added? Well that was the AI agent operator. So what Alistair was explaining to you there is what a tech AI agent operator looks like and that's the nature of the job they don't run in tech agents but the interesting thing in the report was actually the second or third amount of different jobs that they were adding was around AI agent operators and other verticals of the business as well it's not just in tech now those guys in tech have led a lot of the the use of these agents were sort of naturally gonna be the forefront to some of these things but also note that not

Not everyone in tech was a fan of this idea, right? There was loads of people, even this time in the middle of last year, saying this AI stuff just can't do the coding. It's not that good. It's not going to be good enough. It's not going to be as good as us human developers in this stuff.

And those same commentators are now, a year and a bit later, just going, no, actually, this stuff is great. And now it really has flipped, and it's going to be flipping in other areas of business. Everything's going to be changing.

And that doesn't mean everyone loses their job, and we're definitely not seeing that at all, actually. AI has net created jobs in the US over the last year. It's definitely the stats are coming out now. There's not this huge apocalypse of layoffs that we're seeing.

From Coding Agents to Marketing Systems

So what I'm going to talk about a little bit more today is how these AI agent tools and AI agent systems begin to work in other areas. Okay, so I'm going to talk about marketing, but hopefully that will let you give an insight outside of the tech and the code side of things of what happens in other areas of the business. because any given business is one is building a product or service the other part of the business is marketing or selling that product or service that's the essence of business right so this is the marketing side of this stuff so what

does agent operators look like well the other good bit about this is I definitely believe in humans as a member and an orchestrator of these systems I am NOT a I there's definitely some guys in the AI tech community and say agents are gonna just take care of everything we're gonna be able to automate the the world and I definitely am not a subscriber to that at all, partly because

even with cloud design and the look looks a bit cloudy is actually a thing right, it really does. I could point out the technical bits of why Andy's slides looked a bit cloudy with a capital letter eyebrow over a newsreader font mechanic and a subtitle in a sans serif font. That's what makes it all look a little bit cloudy and that looks a bit boring very suddenly and it's only been out about four months and everyone can now see it very clearly in various things, it's quite interesting how good humans are at that pattern recognition.

Why Human Judgment Still Matters

So what's important in marketing is judgment and taste and brand and direction, those are the things that at a high level a marketer wants to do or a business owner wants to do when they're selling their product, they want it to look like them in many cases and there's a high quality standard.

So what we've done really from a few years ago of just question and answer back with initial chat GPT, that was just ask a question that gives you an answer turn this into a two -way conversation.

What we've now got as agents, as these guys have been getting to talk about, Alistair definitely pointed to the idea of an agentic harness that does a lot more things.

So Claude and ChatGPT still look very much like it was when they first released Under the Hood. They're very, very, very different actually in what they're actually doing. And what they're doing is this agentic harness work that Alistair was alluding to.

But you can do quite a lot of this stuff just with ChatGPT and I think that's a big factor in this.

Building a Marketing Agent Harness

I can write an email with ChatGPT or Claude or Gemini. I can create imagery with any of those nano banana from Google is very very good for that. Create landing pages, that's all pretty easy to do but all these things aren't talking, you're copying and pasting from a number of different applications but there's other jobs that need to be done in this process as well. So what

What we're beginning to see are vertical -based harnesses coming in. They do specific jobs for businesses, not just – there's a couple of ways the world can go, and we don't know – and to be honest, to be open, I'm not sure which way it's going to go.

It could be ChatGPT and Cloud and Anthropic will be the dominant companies in the world, and every other bit of software is going to go away, and it's just going to be two pieces of software. I sort of don't believe that, just in terms of it feels a bit too far -fetched as well. They're very capable, but I don't think they're going to do everything.

that's sort of the mechanics and part of that is because people have other jobs to do that we're not seeing those things do so let me go through a little bit of what that begins to look like so what does a marketing harness need to know different from another thing so what are the different jobs that need to be done

The Context Marketing Agents Need

really Alistair talked about giving it the background the knowledge of how your product project works overview these sorts of things and really in marketing marketing what is that what's how does our marketing work it's a strategy here's my strategy here's the overall view of that how do we structure that and put it in there now we can sit there and

do this with claude or in skills etc if you're pretty good and very comfortable with adding all that stuff you can begin to add that stuff in there as well so that's possible who's my audience who's my ideal customer profile as alistair was talking about in there who's the person or people i want to sell to this but who's my ideal customer okay but when what does my brand look like from

that strategy and customer you build your brand out from that the colors the fonts your tone of voice etc what you do say what you don't say what's your point of view as a brand these things are important again you can put all this into the back of cod and it will begin to do some stuff but really what we want to be able to do with that background with that information is then to say

Turning Strategy into Campaigns

okay now i've got a target i need to get an extra a thousand visitors to my website to do that i I want to do social campaigns across TikTok, LinkedIn, Instagram. I've got this extra offer that we're doing, and that's what we want to do.

And I want to maybe create a landing page for those people who can click on from those with maybe a short link that brings me back to a landing page where people can sign up for my offer or whatever or buy my product online.

So that is like the whole, that's actually the real job that you need to do, not just write an email or create a social post. You actually want to do all of that stuff, right?

And then from that campaign idea, you get your content, you build it out, and then you want to publish that. And actually, what happens after you do that then in professional marketing departments out there, even just for small business running stuff, I want to see the analytics. I want to see what's actually happened, how many people did visit my website, how many new customers did I get, how many products did I sell. So that's the analytics and how that all pulls back in the results.

and ideally what we're trying to do what we set out to do with the stuff that we're looking at is pull all that stuff together and make it into a coherent system that's beyond with these harnesses around instead of just using individual tools how does that look like if it all comes together so that again

this is the sort of like island concept of this but actually it's trying to take care of the full job as opposed to just bits of it and and that's what all these different elements of the task and that agents can now begin to do and the role of the person now defining this now begins to change, okay?

The Human as Agent Orchestrator

So as opposed to, we've been talking to people who've been looking at our platform going, oh my goodness, this is going to save me one day a week and some of the jobs that they do because it takes care of all the stuff.

What they do is now goes, the interesting bit that we have found when we've seen people using these AI tools more and more, one of the words that they talk about is fun. It's quite interesting. It comes up a lot across a lot of things.

That sounds a bit strange but i know a lot of the developers who are using this are having more fun than they've had in a long time because they're having they're allowed to move up the metal layer they're not worried about some of the specific details they're more guiding and steering they're being more of a commander role in what it's doing they're setting the strategy they're setting the objectives and then they're the ai is doing a lot of the heavy duty lifting but then they're tweaking it generally at the end as well especially when it comes to marketing

there's definitely an element that you need to make sure that it's not producing AI slop what we do want to have is fill the world with AI slop that is not the objective here and that's not good marketing as well right so that's not going to be good for your brand if you're just pumping out crappy stuff like no one really wants to see that it's not going to convert users and therefore in the back of that you need a human overlay and you need that taste and and brand direction and judgment are really key parts of what these AI agent operators do.

So what they're really doing in these systems is the loop and the analysis of all this different stuff. We're doing the publishing jobs, we're measuring, we're using the analytics and the learning that wrap all of that stuff in together as well. And really, that's the idea.

A Practical Marketing Workflow

So what does that sort of look like? Okay, so what I'm gonna do is give you a couple of screenshots from an app to give you an idea of what that begins to look like in terms of the editing mechanic. mechanic.

So this looks like some other tools you might do for social media planning. You come in at the start of the week, what are we going to do for our marketing for this week and for this month? You put it all together and here's a plan for that. How do you do that? How do you create that quickly?

Well, we have AI tools to help you pull that plan together based on your strategy, what you're trying to do, who you're trying to talk to and what channels you want to go down. So we do it across all these different channels. And with that, you can click on one of these when it comes out. Here's an example of the the post that I put out yesterday, LinkedIn advertising this event tonight, and we get all the data from it. We publish from here, you get all the data coming back in, so the knowledge begins to live within the system.

In theory, some of Claude could maybe go out and pull some of these things. In reality, this is pretty hard to wire up, even if you have the capabilities and skills in Claude Code or ChatGPT, and this is actually what you need to see.

In reality with that, when you get that email coming out, what's it going going to look like you want to be able to edit it natively right you actually generally vibing stuff out the way that Andy showed it to start with cloud design you can get stuff that looks pretty good and you're going to get 80 % of the way 90 % of the way really really fast but in reality

there's a load of little mini tweaks and for my a lot of experience in vibing websites and doing design through just the chat tools it takes quite a long time to get everything exactly right and really good marketers and really good brand people want everything to be exactly right they want exact imagery right, they want exact fonts, the right language exactly there, moving around a little bit.

You want to be able to do that manually and that's what our set of tools lets you do.

Suri, what are you doing man? Okay, Suri's picking up something there, there we go, okay, so there we go.

The Promise of Agentic Systems

So what we're really looking at here, and this is sort of like, this is sort of optimistic for the humanity in this AI world at the moment, is really, you go go from a content generation set of tools and micro individual actions into a sort of system that's sort of like an overall system -based approach that really helps you get your job done in a much more elegant, systematic way. That's really the promise of these agentic harnesses.

And really it takes it like the agents are doing a lot of the work and you're having to sort of become an orchestrator of agents, but that sounds maybe potentially boring, but it's sort of interesting. Your creativity can really begin to flow. You don't need to sweat the details. You're going, this is where we want to go. This is the vision we want to create. You'd be able to do it in a much more natural sort of way, and it does feel more fun. It is quite interesting, that word, but it does feel it's quite addictive in the way that it does stuff, and that's where it comes out.

And basically, this is the offer that we have in that, but that's how these agentic systems are generally working. I've used this as an example, but I think that's what we're going to see in lots of different areas and jobs to be done. These agentic systems Potentially Claude and ChatGPT will be able to do quite a few of these tasks, but we're going to see, I think, some of these agentic harnesses begin to emerge as well alongside that as well.

Try the Platform and Share Feedback

Okay, so here you go. If you want to try out what we've got, I've got like 300 credits on offer for this. This is a QR code, again, created through our system itself. So this is our own links that we're using here as well. So if you want to check it out, go along and have a play.

We're just launching this at the moment. It's probably still a little bit buggy in a few places, but basically check it out. We think it's very good and it'll be interesting to get your feedback.

So please drop me an email as well We'll get it on the list and contacts and be very happy to answer your questions and give me any feedback What do you think good or bad especially bad and we'll try to improve it. That would be great.

Thank you very much