How is AI reshaping agency economics? 

Anne Stagg
CEO
Digitas UK  

When people talk about AI and agency economics, the conversation often centres on efficiency. While that's certainly part of the story, I think the more interesting shift is in where value is being created. 

AI is helping agencies deliver certain tasks more efficiently. Whether that's accelerating research, generating ideas, simplifying processes or making insight more accessible, it's creating opportunities for people to spend more time on the work that requires judgement, creativity and collaboration. But it's simplistic to view that purely as a cost-saving story. As routine processes become faster and more scalable, clients increasingly place greater value on strategic thinking, creativity and the ability to solve complex business challenges. 

At Digitas, we've been exploring how those changes are reflected in the way we operate. We've brought different disciplines closer together and continue to evolve how we build teams around client challenges. AI has an important role to play in that, helping people work more effectively and respond with greater speed and flexibility where needed. 

For me, the most exciting aspect isn't the technology itself, but the opportunity it creates.

The agencies that succeed will be those that combine AI with human expertise to deliver better outcomes for clients and unlock new ways of creating value.  

Rob Baldwin
Northern Europe CFO
dentsu 

Our experience is that most of the discussion assumes AI's principal economic function is to reduce costs. That framing is too narrow, and carries real risk when marketing is under pressure to be more results driven than ever.  

We are seeing an expectation from procurement teams that ‘AI efficiencies’ be reflected as fee reductions. Whilst we understand the motive and imperative, this risks missing the real value that agencies deliver to clients. Price-driven objectives underplay the investment agencies are continually making into new technology, and they cast agencies as an ‘efficiency’ partner rather than an outcomes or ‘effectiveness’ partner, which serves neither party well in the end.  

A more useful distinction is between cost reduction and capability extension. For example, audience definition, decisioning and measurement are established agency competencies. AI's contribution is to extend them; richer audiences, faster decisioning, and sophisticated measurement that operates closer to real time. This is value creation rather than cost extraction, and its returns are of a different order: a one-off saving is finite, whereas a capability that improves client performance will compound over time. The value we bring to clients should not only be measured on the outcomes that campaigns create but the multiples that are generated helping clients to build and access deeper capability. In practice, this is the AI x PI shift: AI-enabled workflows scaling capability, with people providing direction, governance and judgement.  

The question of platform dependency is a legitimate one – and the response is architectural: remaining open and interoperable. We need to regard AI model providers – or those that host AI models, such as LLMs - as suppliers rather than as proprietors of the agency's value. That value resides in orchestration, workflow design, data and human judgement. This is the layer around the AI model, not the model itself.  

As suppliers, model providers will in time compete on price as well as capability. Particularly as agencies recalibrate efforts internally to understand which models best ‘fit’ given tasks. If an agency can preserve genuine choice and observe what each model costs and delivers, they should remain an affordable and net-positive tool rather than an uncontrolled cost. The same logic applies commercially: outcome-linked models should be used selectively, where measurement, attribution and agency influence are credible.  

Whilst we will continue to seek efficiency, we believe the bigger economic question is not how quickly costs can be reduced, but how the industry chooses to value what AI creates.  

Liz Duff
Managing Director
Mediaplus UK  

The conversation around AI in our industry tends to focus on efficiency gains. I think there's a bigger question we're avoiding: are we being honest about what AI actually changes commercially, or are we reaching for a story we've been telling ourselves for twenty years? 

Value-based pricing has been the industry's aspiration since long before AI arrived, and it hasn't taken hold at scale. Outcomes are hard to attribute cleanly to any one input, and clients have generally resisted paying for results they can't fully audit. AI doesn't solve that problem. If anything, it makes it harder to separate what the technology contributed from what the team did. Agencies claiming a clean pivot to outcome-based fees should be honest about how much of that is capability and how much is ambition. 

Our own approach is to stop trying to reprice speed. Where AI genuinely reduces the time and cost of production and execution, we'd rather pass that saving on than dress it up as new value.

Margin, in our view, should sit where scarcity still genuinely exists: in strategic judgement, negotiation and accountability, the things AI doesn't do for us.

That's a more honest place to make money than pricing efficiency gains as if they were insight. 

There's also a less comfortable dynamic at play. We've been here before. Agencies spent years building strategy and investment around a small number of dominant platforms, only to find ourselves price-takers when those platforms changed their terms. We're now doing something similar with AI: concentrating capability, workflow and cost around a handful of providers whose pricing and product decisions sit entirely outside our control. That's not a reason to slow adoption, but it is a reason to treat vendor dependency as a genuine commercial risk, not a footnote. 

None of this means AI can't reshape agency economics for the better. It means the industry needs to be more rigorous, not more optimistic. The agencies that benefit will be those that can genuinely evidence the value AI helps create, price accordingly, and stay clear-eyed about how much control they're handing to a small number of technology providers along the way. 

Rob Grealis
CFO
Buymedia  

AI helps teams work smarter, using their experience and time more efficiently and ultimately delivering smarter and more effective campaigns. AI is reshaping media buying and the shift is already happening. According to Omnicom, 80% of all media buying will be AI influenced by 2027 but there’s still a reluctance for agencies to incorporate, despite 44.1% citing inefficient processes as their biggest challenge.  

Using data and insights to make informed decisions, stepping away from bias decision making and being able to pinpoint your ideal customer, linking in to the brands KPIs and being able to track the media spend, see who converts, why they did, where they are, and what messaging works, is the competitive advantage. 

By leveraging AI to develop your AI persona for precision targeting and interrogation informs the omnichannel investment strategy. Having immediate access to reporting across all media channels through a personalised dashboard allows a single view of your entire media mix so in terms of creating reports for the business there’s no more fragmented dashboards, everything is visible. Add to that the econometric modelling (MMM) inputs into predictive analytics to forecast against business KPIs, it ultimately drives smarter planning investment and channel strategies resulting in increased client retention, improved RoI on advertising investment leading to client revenue growth and  internal efficiencies for clients which delivers cost savings.   

Becky Dainter
Finance & Commercial Director
WAA Chosen   

We've built bespoke internal tools that have supported us moving away from off-the-shelf software subscriptions, cutting licence costs and the administrative overhead of running multiple systems. It's also improved the quality and consistency of our internal processing, because tools built around our own way of working outperform generic alternatives. 

To keep track of the financial impact, we've created dedicated AI nominal codes within our P&L, sitting inside direct labour. This is giving us real visibility into how AI is reshaping the make-up of our labour costs. Where we're still developing is turning that visibility into a proper ROI measure, and working out how those efficiencies should flow through into client quotes rather than simply being absorbed. 

The commercial upside isn't only about cost. AI lets us process large datasets and turn them into insight far faster than before, and it's proving useful in client conversations too. We can prototype ideas and demonstrate concepts during the sell-in stage without committing significant investment upfront. That's opened up opportunities to do more within fixed client budgets, effectively expanding what a given fee can deliver. 

The productivity question is more complicated than it looks.

Some of our most AI-literate people, particularly in digital, are producing what looks like a four-to-one output improvement.

But we haven't seen that translate into sales or profitability yet, and that's the harder question; is this genuine productivity, or people simply doing more within the same fee? Until we can measure and monetise it properly, it's a live issue for how we price and charge clients, not a solved one. It's also improved the accuracy and timeliness of timesheets, by giving us better visibility of how time is genuinely being spent. 

Alex Watson
Managing Director
Arke Agency  

AI hasn't changed our business model, it's changed what's possible within our operating model.

We still charge for expertise and time, but what that time delivers has changed dramatically. Tasks that once took a day can now be completed in a fraction of the time. 

The key is that within specific guardrails and AI policies, it can create the first draft, never the finished product. The value still comes from our team's judgement, experience and quality assurance. That's why we haven't embedded AI into our client processes as standard. Every tool is tested thoroughly before it reaches live work. 

A lot of that work sits with me. As Managing Director, I'm investing time into briefing, building and testing AI tools/apps that solve real operational and client challenges. We review every tool on the same criteria: does it save time, is it easy to use, do people actually adopt it, and does it create enough value to justify the investment? 

The investment is primarily in designing and building the tools. Once they're in place, the workflows, logic and decision-making models are already embedded, meaning future analysis and production only require a relatively small amount of AI “credit”. That creates a continued scalable model with low ongoing operating costs, while allowing the tools to continually evolve as our business and clients' needs change. 

What AI gives us is choice. We can reduce the cost of delivery, provide greater depth for the same budget, or deliver work faster. The right answer depends on the client and where they'll see the greatest return. 

The bigger opportunity is deciding what happens to the tools we build. Some stay internal, some become client-facing, and others have the potential to become products in their own right. 

The biggest challenge isn't building the technology. It's making sure what we create is genuinely practical, that people actually use it, and that it evolves alongside our business and our clients' needs. A clever tool that nobody adopts has a short lifespan and little value. 

AI isn't replacing agency value. It's giving agencies more ways to add value. The advantage comes from knowing what to build, how to use it, and where it creates the greatest commercial impact.   

Adam Cleaver
Founding Partner
Collective    

The productivity story around AI gets most of the attention, but from where I sit, the more revealing question is a financial one: what does the output actually cost us once you follow it all the way through? 

What we’ve found is that the headline efficiency rarely survives contact with reality. Generative tools produce plausible work fast - but plausible isn't the same as on-brand, accurate, or consistent, and the gap between the two lands on someone's desk to fix. When we've looked honestly at where time goes, a lot of the "saving" from generating something quickly gets spent again catching drift, correcting errors, and reworking assets that were never built to a standard we could reuse.

Our take is that ungoverned AI tends to move cost around rather than take it out, and it adds brand-safety and compliance risk while it does. 

So our approach has been to stop treating this as a creative problem and start treating it as a control problem. We give AI a verified source of truth - a governed twin of the product and brand - that it can't deviate from. The commercial difference is real: output that's right first time, reusable across channels, and auditable when someone asks how it was made. 

It also protects us against a risk I think finance leaders are right to watch - becoming dependent on a few AI platforms whose pricing we don't set. By building on open standards and keeping those governed assets in the client's hands, the value compounds with them. 

Governance, for us, hasn't been a brake on AI. It's been the thing that made the returns real. 

Richard Brooks
Co-founder and Director
Kinase   

The prevailing assumption is AI will make agencies leaner and clients less dependent on them. At Kinase, our experience tells a different story.

AI hasn't reduced what clients need from us, it has increased it.

More strategy but also more of the things that weren't practical before, like integrating performance creative or better measurement solutions. 

That's why the idea that AI can replace staff is ludicrous. It's a powerful tool for our employees, not a replacement, and we're actively investing in recruitment and training because the alternative is a false economy. Even before AI, we saw a shift toward more senior teams driven by client demand for strategy and that trend continues. However, you still need juniors coming in to become the strategists of tomorrow - lose that pipeline and the institutional knowledge goes with it, which is hard to get back. 

Our structure as an Employee-Owned Trust shapes how we think about profitability: efficiency gains get reinvested as a combination of client savings and rewards for the people harnessing new tools most effectively. It has never really been about time spent for us - what matters is outcomes and the best outcomes come from people and technology working together, not chasing short-term headcount savings. 

This logic matters given that AI is proving costly for providers, who are looking for new ways to charge. The assumption that AI is cheap may not last that much longer. People now choose between different AI models and use the one which offers the best value for the task they’re doing. Given the fast pace at which the industry is moving, you certainly can’t put all of your eggs in the AI basket without risking disruption.