Every agency claims to "use AI" now. Most are vague about what that means, which is a strange way to talk about something that touches your brand every day.
So here's the plain version. We've spent the last two years rebuilding how The Blink works around AI. Not as a gimmick in a pitch deck, but as the engine room of the studio. What we learned is simple to say and harder to practise: AI is the accelerator. Humans are the driver. Take the driver out and you don't get a faster car. You get a crash, delivered at scale.
What AI is genuinely good at
Let's give the machines their due. There are parts of marketing work where AI isn't a nice-to-have any more. It's simply the better tool.
- Research at speed. Summarising a market, a competitor set or two hundred customer reviews in minutes, not days.
- First drafts. Headlines, captions, product descriptions, ad variations. Not the final word, but a fast, wide starting point.
- Asset production at scale. Resizing, reformatting, versioning one campaign into forty placements.
- Personalisation. Writing the same message ten ways for ten audiences without ten copywriters.
- Monitoring. Watching ad accounts, search rankings and analytics around the clock, and flagging what moved.
All of these are high-volume, repetitive and pattern-based. That's AI's home turf.
What it's still bad at
And then there's the stuff that actually decides whether marketing works.
- Strategy. Choosing what not to do. AI will happily give you twelve options. It won't tell you which eleven to kill.
- Taste. Knowing that one typeface, one word, one frame is right and the other is merely fine.
- Judgement. Reading a room, a market, a client's nerves on launch day.
- Accountability. When a campaign goes wrong, a model doesn't pick up the phone.
These remain stubbornly, gloriously human. Any agency pretending otherwise is either selling you something or hasn't been paying attention.
How to use AI in marketing (without blending in)
The biggest risk with AI isn't that it gets things wrong. It's that it gets things average. Language models are, by design, very good at producing the most likely next word. The most likely next word is exactly what every one of your competitors is also getting. Average is the enemy of a brand.
Here's how we keep the speed and lose the sameness.
1. Feed it your brand, not the internet's
A blank prompt gives you the internet's voice. Give the model your tone of voice guide, your best-performing posts, your product truths and the words you never use. If you don't have those written down yet, start there. Our guide to finding your tone of voice walks you through it.
2. Use it wide, then go narrow yourself
Ask for thirty headlines, not one. Then a human picks three and rewrites the best one. AI is a superb generator of options. It's a poor judge of them.
3. Keep a human between the machine and the customer
Nothing goes out under your name that a person hasn't read, with their brain switched on. Not because AI always makes mistakes, but because when it does, they're confident, fluent and very public.
4. Measure outcomes, not output
AI makes it trivially easy to produce more. More posts, more ads, more blog articles. More is not a strategy. Judge your AI workflow on the numbers that matter: enquiries, bookings, sales, cost per customer. If those don't move, you've just automated noise. (More on that in Numbers, not likes.)
5. Write your rules before you need them
Decide now what AI may and may not do with your brand, while it's calm. Not mid-crisis.
The lines we don't cross
These are ours. Steal them if they're useful.
- No final copy without human review. AI can draft. Humans approve.
- No client-facing deepfakes. No synthetic people presented as real, no fake testimonials, no manipulated imagery passed off as genuine. Regardless of how much time it might save.
- No replacing your voice. AI writes in your tone, not its own. We train it on your material, then we edit.
Clients sometimes ask whether this slows us down. A little, at the very end. It's also the reason they trust what we ship.
So where does that leave you?
If you're a business owner wondering whether to "do AI", the honest answer is that you already are. Your ad platforms run on it. Your phone's camera runs on it. Your customers ask ChatGPT for recommendations before they ever reach your website (here's how to get recommended by ChatGPT).
The real question is who's driving. Use AI to go faster at the things that don't need you. Spend the time it saves on the things that do: the idea, the edit, the customer.
That's the whole operating model. It's written up in more detail on our Approach page, and it's how we run every discipline we offer.