Think, then prompt: AI and human ingenuity in digital marketing

I was rifling through piles of notes and voice memos from recent industry events, and I’d intended to summarise the lot using Claude and start planning a new article based on that. And then I paused...

What about my own thoughts? My team’s thoughts? The thoughts of the other industry pros I’ve always bounced ideas off in various WhatsApp groups? I’d just sort of... skipped the human feedback loop entirely. Given I’ve been grappling with how to write about AI in a balanced way, this was a great moment of clarity for me.

It’s how I settled on what I wanted to write about next.

About how your AI marketing strategy can only be as good as your thinking.

So before running any notes through AI, I took a walk, thought about what stuck in MY memory, asked my team and network of fellow marketers for their views on certain things, and THEN used AI to help sort everything into something cohesive. Those human guardrails, that honing of the discourse through the prism of professional ingenuity, is still so vital.

Here’s why.

I’ve been in the social media game long enough now to start recognising the cyclical nature of certain things. I remember when I was coming up in the digital marketing industry, learning from veterans about how this new-fangled doohickey and that hyped-up thingamabob were retreads of stuff they’d seen decades prior. Back then, I was a bit sceptical. With hindsight, I can see the patterns starting to emerge myself!

The one I’m thinking about at the moment is platform dependency. Good ol’ all-yer-eggs-in-one-basket.

It was only about 15-ish years ago that brands went all-in on Facebook. Built massive followings, invested serious budget, treated it like owned infrastructure. Then the algorithm shifted, organic reach fell off a cliff, and anyone without an email list was basically starting again from zero. You moved all your furniture in, then Zuckerberg changed the locks, ate the key, then legged it, cackling, into the dark of night...

My Spidey sense is starting to feel the same thing with AI.

Businesses are wiring their content, their processes, their customer interactions, and a whole lot more into platforms they don't control (unless they’re a smart cookie running local models on their own hardware, but that’s a whole other conversation), and the costs are climbing, and the dependency deepening. If pricing becomes untenable or access gets rug-pulled, do you have a Plan B?

It’s a conversation that keeps coming up at events like Silicon Brighton’s EVOLVE. There’s a proper conversation happening about AI sovereignty. That as of mid-2026, the UK doesn't have its own large language models (though some projects are in development, at time of writing), and if the US or China tightens access to theirs, we're gonna be left high and dry.

It's a national-scale version of the same problem. Your business, my agency, the whole country: we're all gonna be left with keys that don’t fit the lock unless we build something we own.

Which is why, tale as old as time, the boring stuff (which is actually pretty exciting if you’ve got the right mindset!) still matters. Your email list. Your owned media. Your direct customer relationships. Manage it right and drop in a big dollop of your own creativity and identity, and you’ve got yourself a trusty lifeboat for these volatile seas.

A developer at EVOLVE said something that put into words another feeling I’d been having: he compared working with AI to having a junior engineer who never quite grows into a senior one.

They get faster. They get better at the task you set them. But they never really learn to question whether the task was right in the first place. They never think about the customer differently, never wander off-brief and stumble back with something nobody asked for but everybody needed. They just do the thing, quicker each time and with increasing (and sometimes impressive) confidence, sure. Almost by nature, there’s no true “out of the box thinking”.

I brought this back to my team and we mulled it over, and concurred that it echoes something we see in marketing a fair bit: AI can produce 40 ad variations before lunchtime. It can write copy, resize creative, generate headlines until the cows come home. What it can't do is look at a strategy and go “hang about, this is aimed at the wrong people."

It CAN do that based on data from prior performance, which is great! But that data needs to be enriched with fresh human perspective, or it’s gonna go stale really fast. Even if you’ve built a cutting-edge tool that looks up every point of context imaginable — the significance of a meme that hit the web at 8am that morning; hyperlocal reactions to nearby goings-on; the weather forecast for 1:13pm next Tuesday — there’s nary an idea that doesn’t benefit from a little human sharpening.

That step can come in the prompt. Or after the prompt. Or both. Ideally, it happens at every stage of campaign development, and then campaign review after the fact. So don’t let AI change the locks. Give it its own key. Move in together. Ask it for its opinions on where to place the furniture. Reject it when it tells you to put the sofa right in front of the electric sockets. “Don’t be daft,” tell it, “that’s where I plug in the vacuum cleaner.” “Oh, sorry!” it will say. “I’ve never done household chores, and also do not have hands.” And thus, the beauty of humans and machines learning from one another.

So where does that leave us as an agency? We sat with this for a while. Our conclusion, in the middle of 2026, doesn’t sound groundbreaking on paper, but we’ve seen clients welcome it like a cool glass of water in a desert of aimless AI-first proposals.

Good idea first. Then meaningful automation. AI as the engine, not the driver.

It’s why, at a time when it felt like almost all digital marketing chatter was about AI, we’ve been all-in on unpacking changing advertising laws and planning for a 2030 dominated by hyperlocal.

AI without those ideas = weak tea (chuck it down the drain).

Those ideas without AI = passable tea (drink it but wince a bit).

Those ideas complemented by AI = perfect brew (let out an “aah, lovely” and nod with pride).

This could all change overnight, of course. Maybe some big breakthrough will drop a model into our laps that nails human reasoning and meme perception with a level of nuance that makes it feel less like a response to a pool of data and more like an emotional, intelligent reaction to something. If that happens, we’ll be ready.

Until then, we’re shrugging off any “AI everything or you’ll get left behind” anxieties that are bandied around, in favour of letting the successes we score for our clients shape the narrative and define our next steps.

Someone who works on the legal side of things in the industry recently told me that law firms are having to COMPLETELY rethink the way they charge.

If AI can draft a contract in minutes, billing eight hours for it with a straight face gets tricky. So some firms are carving out new pricing categories, and one of them was literally called “judgment.” The idea being: you pay a premium for the expert opinion, the contextual read, the professional instinct. The bit the machine can’t do with the same level of trustworthiness.

Again, you stick around in an industry for long enough, you start to the see the cycles...

The creative industry has worked pretty much the same way for decades. When designer Carolyn Davidson developed the now-omnipresent Nike Swoosh in 1971, she invoiced the company just $35 for the work. It’s sort of an open secret that sometimes, a designer will draw the greatest logo of all time in 10 minutes, then spend the rest of their 20 contracted hours coming up with worse concepts to flesh out the slide decks and ensure the good one gets chosen.

You’re not paying for the hours spent on the thing. You’re paying for the years spent honing the craft and expertise necessary to deliver great stuff in a short amount of time.

Now, lawyers, consultants, accountants (industries that have billed by the hour since forever) are having to make the same distinction. Execution vs. insight. Task vs. judgment. What the machine does vs. what only a human with scars and stories can do.

Same thing is happening marketing. Agencies don’t lose out because some elements of campaign development and analysis can be streamlined. It empowers us to spend more time on creative concepts, to take more measured risks based on deeper insights into a wider arsenal of datasets, and to more clearly visualise changing consumer behaviour, so we can make informed bets on the future of the industry, like the prevalence of hyperlocal come 2030.

AI doesn’t know putting the sofa in front of the window means you can’t see your phone screen between 7am and 1pm because of the glare from the sun.

Your AI strategy can only be as good as your thinking.