WPP’s AI chief says ad agents are still in their awkward phase

AI agents are the newest obsession in advertising tech. Every platform seems to have one, every demo promises autonomy, and every pitch deck suggests software is on the verge of handling serious marketing work with limited supervision.

But one of the industry’s biggest agency groups is signaling that the reality is much messier.

WPP’s AI leadership is taking a notably grounded view of the agent boom, arguing that these systems are still early, unpredictable and not yet ready to be treated like fully dependable operators across complex advertising workflows. The framing is blunt, but the core point is familiar: the technology is moving fast, while trust, governance and consistency are lagging behind.

That matters because agencies and brands are now under pressure to decide where agentic tools belong in the day-to-day business of marketing. It is one thing for AI to summarize campaign notes, draft media plans or surface performance anomalies. It is another for it to autonomously make decisions that affect budget allocation, creative output, targeting strategy or client relationships.

In other words, the industry is shifting from experimentation to operational questions.

For holding companies like WPP, the appeal is obvious. Agents could eventually help automate repetitive workflows, connect fragmented systems, speed up planning and make teams more efficient. In a sector where margins are tight and execution is often messy, even small productivity gains can have a big impact.

Still, that upside comes with a familiar problem: a lot of adtech has a tendency to market future potential as present capability.

The caution from WPP lands at a moment when “agentic AI” is becoming the label of the year. In practice, that label can cover a wide range of tools: assistants that follow prompts, bots that chain together tasks, systems that interact with software interfaces, or models that make recommendations with varying degrees of independence.

The problem is that those categories are often blurred together. A tool that automates a narrow task in a controlled setting is not the same thing as a system that can reliably run multi-step campaign operations in the wild. Advertising is full of edge cases, changing priorities, platform quirks and brand sensitivities. That makes it a tough environment for anything claiming meaningful autonomy.

There is also a governance issue. The more authority an agent is given, the more important it becomes to understand what it is doing, why it is doing it, and how quickly a human can intervene when something goes wrong. In media buying and campaign management, mistakes do not stay theoretical for long. They can affect spend, performance and brand safety in real time.

That is why the current phase of adoption looks less like replacement and more like supervised assistance. Agencies are testing where agents can remove friction, accelerate research, draft outputs, manage internal processes or support optimization work. But the stronger the claim of autonomy, the more skepticism tends to follow.

This is not a rejection of the technology. If anything, it is a sign the conversation is getting more serious. The first wave of excitement focused on what agents might become. The next phase is about what they can consistently do now without creating new risks for marketers.

That distinction is especially important in an agency setting, where AI tools have to work across different clients, objectives, compliance demands and tech stacks. A flashy prototype may impress in a demo. It is much harder to prove reliability across live campaigns, approval chains and performance accountability.

For brands, the takeaway is pretty simple. Ask less about whether a vendor has an agent, and more about what that agent can actually do under real conditions. Where does it save time? Where does it need oversight? What decisions can it make, and which ones still require a human in the loop?

The hype cycle around AI agents is not ending anytime soon. But if one of the world’s largest ad groups is warning that the technology is still in an awkward developmental phase, marketers would be smart to treat autonomy as a spectrum, not a switch.

That may be the healthiest sign yet that the industry is finally growing up around AI — even if the agents are not there yet.