Where AI earns its place and where it doesn't.

What's the honest frame for AI in marketing?

AI is a genuine force-multiplier in a few specific slots, and a liability in others. That's the whole thing. Anyone telling you it does everything is selling you something, and anyone telling you it does nothing hasn't actually used it properly.

So here's the honest map, where I'd hand work to AI without blinking, where I refuse to, and why. It's less about the tools and more about which parts of the job wear a team down versus which parts make a team special. Get that split right and you get far more out of the people you already have.

Where does AI genuinely earn its place?

On the exhausting stuff nobody wants to do. The actions people can't be bothered taking, not because they're hard, but because they're relentless.

The one that stands out most for me is the year-long orchestration of marketing, the boring project management of an entire year. Think about it as two bookends with the good bit in the middle.

• The front end, planning backwards. Every event, launch, collab and activity coming up, all mapped onto a timeline. Then working back from each one: if a product lands in September, when do you start teasing it? When do the previews go out? When does the pre-order open? When do you turn the hype up? That's a mountain of sequencing, and it reshuffles every time a new date appears.
• The back end, actually shipping. The posting, the delivering, the publishing. The bit where someone's stuck trying to push out too many posts and too many blogs on time, and the dates start slipping.

Both of those bookends are perfect for AI. They're where teams miss timings, mis-prioritise, and quietly leave return on the table, not because they're bad at their jobs, but because keeping a whole year straight in your head is exhausting and thankless.

When you take that off the team, the economics change. You stop missing the days your campaigns were meant to go out. You stop putting the low-impact thing first. You push out far more content, more accurately, and you actually cash in on the plans you already had.

"It lets you put out so much more content, be more accurate with it, and take advantage of all the plans you already had, so you get more back from your existing team. I've seen it internally and with clients."

That's the promise, said plainly: not magic, just more output from the same people, because they're no longer drowning in admin. If you want the fuller picture of how we build these, that's our Marketing AI work.

Where does AI NOT earn its place?

The precious bit in the middle, the genuinely creative marketing. The pin between those two bookends is the actual marketing: the campaign ideas, the newsletter angle, the best creative minds in a room deciding what this thing should be.

That's exactly where your creative people should be spending their energy, and exactly where they can't be if they're buried in "is this more important than that, should this run for three months or one". Hand the project management to AI so the humans get to do the human bit. Don't hand over the human bit.

The one thing I won't hand to AI

Design. Specifically initial design, and I'll die on this hill.

By initial design I mean the core creative DNA: your logos, your brand guidelines, the way a webpage actually looks, how the buttons feel, how the interactions work. AI design looks clean and nice and perfectly fine. It also looks generic, it's built to please everyone, so it feels specific to no one.

"AI design looks clean and fine, but it's generic, built to please everyone. A real human creative makes something extraordinary and specific to your brand, and that's still the centre of the whole thing."

Here's the nuance I really want you to take away, though, because "never use AI for design" would be lazy and wrong. I will hand over a design system once a human has built it.

Picture a website. A human designer builds you a stunning homepage and maybe two or three core pages, and bakes that into a proper design system, the blocks, the rules, the look. That part is human, and it's genuinely unlike anything else out there. Then AI takes those building blocks and builds out the other 80 pages you don't obsess over but still need, for depth, for engagement, for AEO and search.

On those pages AI lands 90-95% of what you wanted, because it's working from your components, not inventing from scratch. It won't scare anyone who arrives there. It still reads as your brand. And ideally they'll wander to your pristine homepage and see the real, extraordinary version. Human owns the creative core; AI scales it. Both jobs done by the thing that's best at them.

What does the human-in-the-loop pattern look like?

It looks like a tightly-scoped orchestrator with a human sitting right in the middle of it, not one clever agent trying to do everything.

The single most useful thing we've shipped is exactly that: an orchestrator for our own marketing. It doesn't write, it doesn't design, it doesn't generate ideas. It has one job. Keep every event happening in the business front of mind, and tell us what needs to start when.

It pulls new events out of meetings and emails, drops them onto the timeline, and cross-references our own data on what's actually worked with past launches and collabs, so priority is based on evidence, not on whatever felt urgent that morning. Every week it's current. When a launch shifts from September to October, the whole plan reshuffles itself instead of rotting in a spreadsheet.

Crucially, we trust it because of what it doesn't do. It never speaks to a customer. It talks to us, "you're launching this on that day, so you need to start this now", and it tells the other agents when to begin their jobs. That's the pattern worth copying:

• Scope tightly. One agent, one job. An orchestrator that also tries to design is an orchestrator you can't trust.
• Keep a human on the creative and the call. The agent surfaces and sequences; a person decides.
• Point the risky, brand-facing output inward first. Internal instructions are safe to automate. Customer-facing voice earns trust slowly.

That's the split we run across the tools. ChatGPT and Claude for the thinking, n8n for the plumbing, Airtable and Notion for the timeline everything hangs off.

How do you pilot it without betting the brand?

Start on the boring bookends, keep it internal, and only widen the scope once trust is earned.

Don't launch by pointing an agent at your customers on day one. Pick the exhausting, low-risk admin first, the year plan, the scheduling, the publishing queue. Let AI take that off your team, watch the output climb, and keep a human reviewing anything that leaves the building. Protect the creative core, campaigns, brand, initial design, as human work from the start, and only hand AI the scaling once a person has set the blocks.

That's how you get the upside without the downside: more content, better timing, more return from plans you already had, and a brand that still sounds and looks like you. If you want a hand mapping which slots are safe to automate first, request a quote and we'll walk your calendar with you.