The AI Gap in Agencies Is Getting Real (and Most People Are Still “Trialling”)

05/02/2026

Peter Czapp has been spotting something interesting (and slightly depressing, if you run an agency): the gap between agencies who actually use AI well… and everyone else who’s sort of poking it with a stick.

He’s been digging into early results from BenchPress 2026 (apparently “no one in the world has seen these results yet”, which is exactly the kind of thing marketers say right before they hit you with a chart).

Here’s the punch in the face.

What agencies are experiencing right now

From the early data:

  • Nearly 40% of agencies have experienced clients taking work in-house

  • A third say clients expect to pay less

  • For many teams, AI has been a distraction without delivering value

Which basically translates to: AI is everywhere, but a lot of agencies aren’t getting paid more for it. They’re just getting more noise.

And for most agencies, AI hasn’t been the advantage they hoped it would be. It’s just another thing to keep up with. Another tab open. Another “we should probably do something with this” meeting.

Who’s doing what with AI?

The results split agencies into a few camps:

  • 20% are still in the “trialling” phase (dabbling, experimenting, mildly panicking)

  • 68% have moved beyond that and are finding wins by automating tasks in parts of their business

So most agencies are using AI somewhere. A few automations here, a helpful prompt there, maybe a fancy new workflow for content or reporting.

But here’s the bit that matters.

The tiny group actually winning with AI

According to Peter, only 12% of agencies are really maximising AI’s potential by using it successfully across their entire business.

Not “we use ChatGPT sometimes.”
Not “our strategist has a prompt library.”
Not “we made a GPT for blog outlines.”

Across the whole business.

And for that small group, the impact is extremely positive.

They’re delivering more value, and their cost of delivery has reduced.

Translation: they’re getting faster, cleaner, more scalable, and probably more profitable, while everyone else is arguing about what tool to pick.

My take: AI adoption isn’t the goal. Advantage is.

Most agencies aren’t losing because they didn’t buy the right AI tool.

They’re losing because they haven’t done the boring work:

  • updating delivery systems

  • redesigning processes

  • training the team properly

  • setting standards

  • integrating AI into ops, sales, delivery, QA, reporting, and management

AI doesn’t reward dabbling. It rewards integration.

The agencies pulling ahead are treating AI like electricity. It’s not a department, it’s infrastructure.

What to do if you’re an agency owner reading this

If you want to move from “trialling” to actually benefiting, focus on a few things.

Pick one workflow and industrialise it

Not “use AI more”. Choose something concrete like:

  • briefs to first draft to QA to delivery

  • reporting to insights to recommendations

  • lead research to outreach personalisation to follow-up

Then build the system around it.

Stop relying on the one AI person

If only one person knows how to use AI properly, you don’t have capability. You have a single point of failure.

Make it repeatable. Document it. Train it. Make it normal.

Measure the boring metrics

Track things like:

  • time saved

  • margin improvement

  • turnaround speed

  • revision reduction

  • delivery consistency

If you can’t measure the change, you’re just doing vibes-based transformation. And vibes don’t scale.

The uncomfortable truth

Clients taking work in-house and expecting to pay less is not a temporary phase.

AI is pushing agencies into a split:

  • those who become leaner and more valuable

  • those who become more expensive middlemen

It’s not “AI will replace agencies.”
It’s “agencies who use AI properly will replace agencies who don’t.”

Back to News