In this opinion piece, Sangeeta Leach, seasoned creative agency exec and founding director of The Leach Partnership, unpacks the very real threat AI poses to agency margins and why most creative shops won’t realise all the money has gone before it’s too late. Fortunately, there are things you can do about it.
AI isn’t going to kill agencies.
It’s going to quietly shrink them.
And most won’t notice until the margins are already gone.
I saw a glimpse of this shift myself recently. Over the weekend I experimented with vibe coding – using AI tools to build something without any coding experience. By Monday I had built a few things including a simple margin risk diagnostic for agencies.
Nothing revolutionary, but it made the point clearly. Work that once required specialists, time and budget can now be put together surprisingly quickly by someone willing to experiment.
That change in speed is starting to ripple through agency economics.
For years, agencies have been told to stop charging for time and start charging for value. Most agreed in principle. Few actually made the shift. Now AI is stress testing that gap.
Copy that took days now takes hours. Decks come together faster. Research that once took days can be pulled together in minutes. Production moves at a completely different pace.
Time is no longer scarce in the same way. And if time is not scarce, it is a weak thing to sell.
We are already seeing the shift in how some clients approach agencies. One agency recently walked into a briefing where the client had spent the weekend generating campaign ideas and visuals themselves using tools like Midjourney and Pomelli. The strategic thinking behind them was thin, which is why the agency had been called in. But the assumption had quietly changed. If the client could produce campaign outputs that quickly, surely the agency could too.
The squeeze will not arrive all at once. It will come through small conversations.
“Can we tighten this scope?” “Can we sharpen the fee a little?”
Margins thin quietly and consistently. And by the time the P&L shows the damage, the habit is already set.
The deeper problem is that most agencies do not actually know what commercial value they create. They deliver campaigns. They drive engagement. They win awards. But can they show that the work moves growth, revenue, or market share in numbers a client can take to their CFO or board? Because if you cannot define the value, you cannot defend it.
In almost every agency conversation we have, this is where things get uncomfortable. Not because people don’t care about the answer, but because the model was never built to produce one. Fees were scoped around activity. Value was assumed, not measured. And for a long time, clients didn’t push hard enough to expose it. That time is ending.
Then there is the structural problem.
The agency model runs on layers. Juniors do the groundwork. Mid-level staff build it out. Senior people steer. Margin lives in that pyramid. But AI is very good at the kinds of work that used to sit at the bottom of that structure. Research, first drafts, resizing, versioning and presentations.
One agency we work with recently ran a simple trial. For a few weeks they ran parts of their admin and research work in parallel, the usual human process on one side and a set of Perplexity agents doing the same tasks on the other. The outcome was straightforward. The agents were cheaper and faster for a surprising amount of the work. No one was fired, but it made something clear. The economics of the junior layer had shifted.
Some agencies are already working this out in more fundamental ways.
Most agencies are still treating AI as a production shortcut. I saw an ad from Special Group hiring people whose entire job is to sit inside the creative team and think natively in machine intelligence. Not a tech team. Not a separate capability. A creative role, reimagined from the inside out.
Globally, Monks have made a more structural call. They are moving away from the billable hour and replacing it with a subscription model where clients pay a recurring fee for ongoing access to senior talent, AI workflows and institutional knowledge. The model is expected to account for around 25 per cent of their revenue by end of 2026.
Rather than renegotiating scope every time AI improves their output, the service simply gets better. The client pays for the capability, not the hours. That is a fundamentally different commercial conversation.
The agencies that thrive will not be the ones with the flashiest AI demos. They will be the ones who tie their fees to measurable commercial outcomes, build leaner teams around speed and focus, and get very clear about what they are genuinely indispensable for. Then stop trying to be everything to everyone.
AI is not the threat.
Clinging to a model built on time, layers and the assumption of scarcity is the real risk. Agencies that move now will protect their margins, sharpen their teams and anchor themselves to work clients genuinely cannot do without.
Those that wait will feel the squeeze long before they understand what caused it. By then the margin will already be gone.
