CLARIO / AGENCY WHITE PAPER
DO AGENCIES REALLY NEED MORE AI?
Probably. But that’s probably not the right question.
THE QUESTION IS NO LONGER WHERE AI CAN BE USED.
It is where AI can change the economics of the agency enough to matter.
Every agency CEO I talk to is having some version of the same conversation right now.
Should everyone be on ChatGPT or Claude? What should we automate? Do we need agents? Why isn’t Creative moving faster? What’s the agency down the street doing that we’re not?
I get the urgency. In 2026, Forrester found that nine out of ten U.S. marketing agencies were already using generative AI, for research, concepts, decks, reports, copy, production, all of it. Clients have noticed, and it’s already changing what they expect from you, and what they think they should be paying.
But after looking closely at how agencies actually operate, I think many of us are having this conversation in the wrong order.
Here’s a normal Tuesday.
A client sends over something that’s supposed to be a brief. It isn’t, not really. Accounts fills in the gaps because the deadline’s tomorrow. Creative starts anyway. Halfway through, a call with the client reveals they meant something completely different. The work comes back. Then it comes back again.
Nobody calls this a crisis. The team just gets it done. That’s what good agencies do.
Getting it done, every time, quietly hides how much it’s actually costing you.
The account director spent four hours solving a problem that should never have reached her. Creative built three concepts that were never going to see daylight. A senior person got pulled in because the deadline was suddenly at risk. The ask list kept growing, but nobody stopped to reprice it because having that conversation felt harder than just doing the work. The client got what they needed. Everyone moved on. Almost none of it shows up on the P&L.
Now multiply that by 30 clients. 15 people. 12 months.
That’s where the real AI conversation starts: with the leakage itself.
Agencies are experts at hiding their own inefficiency
This is the strange part about the agency model: being resourceful is the product. Your best people solve things. They stay late, shuffle resources, call the client themselves, rewrite the brief nobody wrote properly, pull in someone senior, negotiate internally, and somehow the work gets across the line. They’re proud of it. They should be.
Clients feel the responsiveness. Leadership sees delivery. What nobody sees is where the cost actually landed.
What nobody sees is where the cost actually landed
We saw this clearly in a recent agency diagnosis. Strong people, a real client-first culture. And that culture was quietly burying the business’s biggest operational problems. Creative was rebuilding the same concept three times because feedback arrived one fragment at a time. Extra requests got absorbed before anyone decided whether to price them. Senior people were still cleaning up exceptions that should have stopped happening months ago.
None of it looked like a five-alarm fire. Together, it added up to a business that delivers brilliantly while making it nearly impossible to see what that delivery actually costs.
It’s not an isolated case. A 2025 study of U.S. marketing and advertising agencies found that 57% were losing between $1,000 and $5,000 a month to scope creep alone. A separate industry survey found that while 59% of agencies grew revenue last year, only 31% improved their margins. Growth and healthy economics are clearly not the same thing.
Now add AI to that agency. You can write the extra concepts faster. Summarize the vague brief faster. Answer the unnecessary client request faster.
You can produce the report nobody reads faster. But is the agency actually better off?
Sometimes AI is the answer. Sometimes the answer is: stop doing that.
I’ve become skeptical of AI workshops that start with ‘where could we use this?’ Give a sharp agency team a whiteboard and two hours and they’ll find forty places AI could be used. Most of the ideas will sound perfectly reasonable. But ‘possible’ and ‘worth investing in’ are very different standards.
If your account team burns hours every month building a client report, sure, an AI agent might cut that dramatically. But I want to know why the report takes hours in the first place. Does the client even read it? Does that information already exist somewhere else in the business? Or is the real issue reporting itself, or that nobody’s ever had the conversation about what the client actually needs to see?
The right fix might be an AI agent. It might be a one-page dashboard. Or a new process. Or someone finally having an uncomfortable conversation about pricing.
Same with creative revisions. AI can compare each new round against the brief and flag when a request has drifted from what was agreed. Genuinely useful. But if everyone already knows revision three always turns into revision six, the real issue is a management decision nobody’s made yet.
Once you stop treating every inefficiency as an AI opportunity, the actual AI opportunities get a lot easier to see.
Start with the business, not the tech
Look for where the agency is spending without realizing it. The strategist still rewriting weak briefs every week. The account team quietly absorbing requests because raising the price feels harder than just doing the work.
Most agencies can tell you exactly how much they invoice a client. Far fewer can tell you what that client really costs.
Most agencies can tell you exactly how much they invoice a client. Far fewer can tell you what that client really costs once you count senior intervention, repeated revisions, unpriced requests, internal coordination and work that should never have happened in the first place.
That is the number I want before I start automating anything.
This is where instinct needs to become economics. You need to know what is happening, how often it happens and what it is costing you.
Once the cost is visible, the solution usually becomes much clearer. Some work should disappear entirely. Some problems need a better process or a management decision. And a smaller group represents genuine opportunities for AI. That last group is where I would invest.
Make AI pay for AI
The first AI implementation inside an agency doesn’t need to impress anyone. It needs to work.
If it removes hundreds of hours of repetitive work, or recovers revenue you were quietly giving away in unbilled work, or frees your best people to spend more time on the work clients actually value, that’s already a result. And results compound: find one source of friction, quantify it, fix it, measure it. If the first intervention saves $80,000 a year, use part of that value to fund the next one. AI stops being a line item leadership has to defend every budget cycle and starts paying for its own expansion.
And recovered time is only valuable if you do something with it. Saving 500 hours does not automatically mean saving the cost of 500 hours. The financial value appears when that capacity replaces future hiring, reduces external costs, allows the agency to take on more work, or moves expensive talent toward work clients are actually willing to pay for. That is why AI opportunities should be measured in business impact, not hours saved.
The conversation changes from ‘we launched three AI agents’ to ‘we recovered capacity, reduced unpriced work and removed a recurring bottleneck.’
That is a conversation a CFO can actually use.
Procurement is already changing the conversation
While agencies are still deciding how to use AI internally, clients are already asking what that efficiency should mean for fees.
Seven in ten marketing procurement leaders say AI has changed their conversations with agencies, and more than a third have already started adjusting remuneration models. Nearly one in three agencies has also been asked for an “AI discount.”
But fewer hours do not automatically mean less value. The harder question is whether the agency can show what its judgment, creativity and expertise are actually producing.
If AI simply means the same deliverable in fewer hours, someone will eventually ask why they are paying the same fee. If it means better decisions, protected margin and more of your best people’s time on work clients truly value, that is a very different conversation.
The agencies that win understand themselves best
AI is becoming a commodity. Your competitor can buy Claude tomorrow, connect the same tools, hire the same people. What they can’t copy is a business that actually knows where it makes money, where it bleeds money, which work clients truly value, and where its best people create disproportionate impact.
That level of clarity starts with diagnosis.
The AI question for agencies is no longer whether to adopt it. The question is where it can change the economics of the business enough to matter. And you cannot answer that question until you know where the economics are breaking today.
ABOUT CLARIO
Clario is a diagnostic built for agencies.
It identifies where time, margin, capacity and opportunity are being lost, quantifies the economic impact, and shows what deserves to be fixed first, including where AI can create the highest return.
Andrea Bovarini and Jo
Co-founders, Clario
SOURCES
1. Forrester, The State Of AI Inside US Marketing Agencies, 2026, released in partnership with the 4A’s, June 24, 2026.
2. Ignition, 2025 Agency Pricing and Cash Flow Report, survey of 273 U.S. agency owners and executives, May 2025.
3. Productive, The State of the Agency Business (2025 Agency Industry Report), survey of 93 agency leaders worldwide, 2025.
4. WFA (World Federation of Advertisers), AI in Marketing Procurement: What Are the Emerging Trends?, benchmark survey of marketing procurement leaders, January 27, 2026.
5. Productive, Agencies in the AI Era: Between Hype and Reality, survey of 180+ agencies worldwide, 2025.