Forrester published a piece of research in November with a title that says more than most headlines manage: "Agentic Advertising's Productivity Paradox." The argument is simple and, once you sit with it, hard to argue against. Agents don't automatically make advertising more productive. According to Forrester, layering AI onto a broken process just automates the inefficiency that was already there, only faster.
That matters for a conversation that's been building quietly among clients and agencies alike. If AI can plan, draft, test and optimize a campaign without a team standing over every step, why should a client keep paying for the hours an agency used to spend doing that manually? It's a fair question. But it assumes the hours were the problem, when often they were the point.
Speed is the easiest expectation to form and the easiest to get wrong. Clients see agentic tools produce several ad variations in the time it used to take a team to produce one, and it's natural to expect pricing and turnaround to shrink in proportion. Forrester's framing pushes back on that assumption directly: the opportunity isn't in doing the old process faster, it's in reengineering the process itself, with people setting strategy and guardrails while agents handle execution and optimization. Those are different jobs, not the same job done quicker.
This is where the pricing conversation gets more complicated than it first appears. If an agency's value used to live in the labor of producing creative, and agents now do a share of that labor, a client might reasonably expect lower fees. But if the agency's real value was always in judgment, in knowing which ad variations are worth testing and which strategy the brand actually needs, that value hasn't gone anywhere. It may just be harder to spot on the invoice.
Quality expectations are shifting too, though not always in a straightforward direction. Agentic tools remove a lot of the tedious groundwork, the reformatting, the tagging, the versioning, and clients can start to assume that whatever comes out the other end must be strategically sound just because it arrived fast. Forrester's point about connected data is useful here. Agents and people both need context to make good decisions, and a tool that moves fast without that context just produces more of the wrong thing, faster.
There's also a question of what gets measured once agents are doing the orchestration. A client used to judging an agency on output, how many ads, how many variations, how quickly a brief turns around, now has to judge something harder to see: whether the guardrails around the agent are sound, and whether the strategy feeding it is the right one. That's a less visible kind of work, and it's reasonable for a client to ask an agency to make it visible.
None of this means clients are wrong to expect change. Agencies that still charge by the hour for tasks agents can do in minutes will have a harder time justifying it, and some of that pressure is overdue. But a client who expects an agency's entire value to collapse into the speed of its tools is making the same mistake Forrester describes: assuming the technology fixes the process rather than running it faster, inefficiencies included.
The more useful question for clients isn't whether an agency uses AI, since most now do in some form. It's what the agency has changed about how it works because of it. Has the process been rebuilt around a clearer division between strategic judgment and automated execution, or has AI just been bolted onto the same workflow that existed five years ago? The first produces work worth paying for. The second produces the same work, faster, at the same risk of being wrong.
For clients and agencies working through this, the practical step is to ask about process before asking about price. An agency that can explain where human judgment sits in its workflow, and where the agent takes over, is giving a clearer answer than one that only promises to be faster.
Sources
- Agentic Advertising’s Productivity Paradox — Forrester, 10 Nov 2026
- The reps were the point — Communicate, 11 Nov 2026
Useful to someone on your team?