Ideate AI does the legwork. People make the calls.

Ideate Partners uses AI every week to read, sort and check the evidence behind your marketing, and helps organisations use it well in their own work. The conversations, the judgement and the decisions stay with people.

Book a free Intro How AI feeds into our work

What does an AI marketing agency do?

An AI marketing agency uses AI to research, plan, make and measure marketing, and helps clients use it well in their own teams. A good one gives AI the reading, sorting and checking, and keeps people on what decides whether marketing works: what customers want, what to say and what to do next.

Where we stand on AI in business

35%
of UK businesses with ten or more staff now use at least one AI technologyONS survey, June 2026, published 20 July 2026
12%
of UK businesses with ten or more staff, in late 2023ONS, Artificial intelligence in UK businesses: 2023 to 2026

AI is good at legwork and unreliable at judgement. So we give it the legwork, and keep the judgement with people you can talk to and hold to account.

Use has grown quickly. Around 35% of UK businesses with ten or more staff now use at least one AI technology, up from around 12% in late 2023. Where firms report something holding them back, a lack of expertise is among the reasons given most often (ONS survey, June 2026, published 20 July 2026).

The tools are easy to get. The value is in knowing what to ask them and how to read what comes back, which is a skill, not a subscription.

We’ve learnt it the practical way, using AI in our own work every week. We’ve learnt the tools hands-on, from chat assistants to agents that hand work on to other agents, and we onboard others on them. That’s the experience behind our consultancy and training.

How AI feeds into our work

Every input is real and written down. Conversations with your customers stay with people.

What we read
- Google Search Console
- Google Analytics
- Our weekly checks
- Competitors’ pages and rankings
- Your customers, in their own words
AI reads and checks
AI from more than one lab
- Claude
- Codex
- ChatGPT
- Gemini
- Copilot
Reads, sorts, compares and flags
People decide
Us
What it means, and what to do
What we make
- Strategy
- Brand
- Websites
- Campaigns
- Reports
All aimed at one thing
Customers choosing you

AI for your business

Four steps, and a person owns every one of them

Anything without a source is flagged for you before it goes anywhere.

01
Gather
AI and automation Your numbers pulled from Google’s own tools every Monday at 7. AI reads the pages that outrank you.
People We talk to your customers and your team, in interviews and workshops.
02
Think
AI and automation AI drafts what the numbers mean and finds the searches where you sit 5th to 20th.
People We weigh the evidence and decide what it means.
03
Make
AI and automation AI helps us build and test, and can help with some first drafts.
People We write and design, to your voice and the facts on file.
04
Check
AI and automation More than 40 checks a week across the sites we look after. Small fixes from a written list, made and logged. AI from a second lab reviews code and client work.
People Anything bigger is written up for us to decide. Then we measure what changed.

The way we work, in numbers

100%
of the 60 checks on our own site ended in a written verdict: Clear, Fixed, Watch or Action needed
Ideate’s weekly write-ups, ideate.partners, 10 Sep to 3 Nov 2026
100
Google results read for every search we track, every week, where Google has that many
Ideate’s weekly tracking, ten Google pages deep, November 2026
2
AI labs on code and client work: one helps make it, the other checks it
Ideate’s second-lab review on code commits, since 13 Sep 2026

Worth knowing

Faster is the wrong goal.
If AI only makes the same marketing quicker, you get more of what wasn’t working, in a voice like everyone else’s. Use it to know more before you decide.
AI earns its keep on legwork, not judgement.
It’s good at reading a lot, sorting it, comparing it and flagging what changed. It’s poor at knowing what a customer meant, what you stand for, or when a number is wrong for a reason it can’t see.
Pilots stall for want of an owner, not a better model.
A trial with no named owner, no measure and no date to decide drifts, then quietly stops. Name the person and the decision first. The tool is the easy part.
One AI checking its own work isn’t a check.
Models from the same lab tend to share blind spots. A model from another lab, then a person, is a real check.

Is a task right for AI? Six questions

Run any task you’re thinking of handing to AI through these, and answer yes or no.

Mostly yes, and it’s a good place to start: try it for a few weeks and measure the time and the quality. A no to either of the last two is a stop until it’s fixed. Mostly no, and AI will probably cost you more time than it saves.

Does the task come round every week, or more often?
Is what it needs already written down, in files or figures?
Could a person check the result in a few minutes?
Would a mistake be caught before a customer saw it?
Can personal and confidential data stay out, or go only into a tool whose terms you’ve read?
Is there one named person who’ll own the result?

Our rules for AI on client work

What we always do

Only what a task needs goes into a tool, and only when it’s needed.
A person signs off the thinking, and everything we write or design for you.
Every claim traces to your own material or a named source. Anything that can’t is flagged for you.
A model from a second lab reviews code and client work before it goes out.

What we never do

Nothing sent to your customers by AI on its own.
No figure, quote or claim made up to fill a gap.
No client’s material used in another client’s work.

The tools that touch client work are Claude, Codex, ChatGPT, Gemini and Copilot. Our privacy notice says how we handle interview notes, which hold personal data.

Built on work that started with people

Questions people ask

Is this a product we buy?

No, there’s nothing to install. The way we use AI comes with the work we do for you. Consultancy, the readiness review, governance, training and automation are services of their own. Our own courses are on the way.

How much does AI marketing cost?

The way we work costs nothing extra: AI is part of how we research, make and check on every project, and the price depends on the work. The services under AI for your business are quoted on their own, once we understand the brief. The first conversation, our free Intro, costs nothing.

Will AI replace my marketing team?

No, and it shouldn’t be asked to. It takes the reading, sorting and checking off people’s desks, which leaves a team, yours or ours, more time for what people do best: talking to customers, deciding what matters and making work people remember.

Is our data safe when you use AI?

Only what a task needs goes into a tool, and only when it’s needed. Anything that identifies a person, interviews included, is personal data under UK GDPR, and we treat it that way. Ask, and we’ll tell you which tools touch your work; if you’d rather one didn’t, it doesn’t.

Which AI tools do you use?

Models from more than one lab, chosen for the task: Claude, Codex, ChatGPT, Gemini and Copilot. Figures come from Google’s own tools wherever we can get them. The tool matters less than what you ask it and how you read the answer.

Where should a small business start with AI?

With one task and one owner, not a tool. Pick something your team does every week, run it through the six questions above, try it for a few weeks and measure the time and the quality. If you’d like a hand choosing, that’s what the readiness review is for.

Can I use ChatGPT for my business?

Yes, with rules agreed first: which plan, what may go in, and who checks what comes out. Plans differ. OpenAI says it doesn’t train on its business plans by default, while on personal plans it may use what you type unless you switch that off. Our guide to choosing AI tools goes through the rest.

Ask us something else