Find your AI champion first

An AI champion is the person inside your team who uses AI on real work, shows everyone else what’s working, and helps them do the same. If you’re trying to get an agency to use AI properly, finding that person is the first thing to do. Before the tools, before the training, before the policy.

I’ve rebuilt four of my own agencies around AI and worked with others as their AI lead. The pattern is the same every time. It sticks when the team owns it.

What an AI champion is

It’s a role, not a job title. A champion keeps doing their normal job. On top of it, they:

• try AI on their own real work first
• show colleagues what worked and what didn’t
• write down the approaches worth repeating
• answer the small questions people are embarrassed to ask
• tell the owners what the team needs: which tools, which rules, what’s getting in the way

You’ll also see it called an AI ambassador or an AI lead. Bigger companies often have a group of them. In an agency of twenty, it’s one person to begin with.

Why telling the team to use AI doesn’t work

The obvious way to roll out AI is from the top. Buy the licences, send the email, run a training day, ask for updates.

It tends to backfire. When people are ordered to use a tool, or measured on how much they use it, they can end up finding ways to look busy with it. On developer forums, people describe use-it-or-else mandates as the thing that killed their motivation. And underneath all of it is the worry nobody says out loud in the meeting: is this how my job goes?

A training day doesn’t fix that either. People enjoy it, go back to their desks, hit the first real problem and return to the way they’ve always worked.

What does work is watching someone you respect do their job faster and better, and then being shown how. People copy a colleague, not a memo.

People copy a colleague they trust, not a memo.

Who makes a good champion

Owners usually reach for the wrong person: the most senior, the most technical, or whoever is loudest about AI. Look for this instead.

• Curious. They’re already trying things in their own time.
• Respected. When they say something works, people believe them.
• Still doing the work. They need real tasks to try it on, and they need to be in the room where the work happens.
• Patient. Half the role is helping people who are nervous or sceptical.
• Sensible about risk. They ask “should we?” as well as “can we?”

They don’t have to be a developer. Some of the best ones are account managers, designers or project managers, because they see where the time goes.

If nobody fits, that’s worth knowing too. It usually means the team doesn’t feel safe experimenting, and that’s the first thing to fix.

What to ask of them

Start with the job they’re sick of doing

Don’t hand them a strategy. Ask them which part of their week they’d most like to never do again: the status report, the first draft, the image resizing, the test plan. Start there.

It’s a good first project because they know exactly what good looks like, they care about getting rid of it, and nobody is upset if the first attempt is rough.

Let everyone else see it working

Give them ten minutes in the team meeting to show what they did. Not a presentation. The real thing, on screen, including the part that went wrong. Showing the rough edges is what makes it credible.

After a few weeks, someone else will ask to try it on their own work. That’s the moment it starts to spread.

Write down what works

Every approach worth repeating goes somewhere the team can find it: the prompt, the steps, what to check before it goes to a client. Over time this becomes your agency’s own way of working with AI, and eventually the standards you write into your agents.

Feed back what the team needs

The champion hears things the owners don’t. Which tool people prefer. Which rule is unclear. Who is worried. Make it easy for them to tell you, and act on it.

What they need from you

This is where it usually goes wrong. The champion is given the title and nothing else.

• Time. Protect some of their week for it. Even half a day makes the difference between a real role and a hobby.
• Cover. Tell the team, and their clients’ account leads, that this is part of their job now.
• Permission to fail. Some experiments won’t work. That’s the point.
• Clear rules. Which tools can touch client data, and what always needs a person to check. They shouldn’t have to guess.
• Recognition. If it changes how the agency works, that should show up in their review and their pay.

When one isn’t enough

Once it’s working, add champions by discipline. A designer learns best from a designer, a developer from a developer. A team of 30 might end up with three or four, meeting for half an hour every couple of weeks to compare notes.

A champion is not a head of AI, though. They make it real inside the team. Somebody still has to set the direction, choose the tools and own the rules. In a small agency that’s usually the owner, often with help from outside. That’s the gap a fractional AI lead fills, and working with your champion is a big part of what I do on the Lead and Embedded plans.

How to start this week

1. Write down the two or three people who are already curious.
2. Ask each of them which job they’d most like to stop doing.
3. Pick one person and one job.
4. Give them the time, and agree what they must not put into an AI tool.
5. Put ten minutes in the next team meeting for them to show what happened.

That’s it. No strategy document, no licences for everyone. One person, one job, and the rest of the team watching.

If you’d like help finding and supporting that person, send me a message.