2.76 out of 5. The grade digital advertising gives its AI.
The players surveyed by IAB Europe rate AI's performance in their operations at 2.76 out of 5. 69% judge it on operational efficiency, and measurement ranks sixth of seven among their concerns. As long as we count hours saved, we won't be able to judge whether it works.
In IAB Europe's 2026 report, 21 respondents rate the performance of AI in their ad operations, "based on the metrics that matter most to you": 2.76 out of 5. The same sample rates its interest in advertising on AI platforms, chatbots and assistants, at 3.54. Among the 19 who answered both, 9 are more interested in the next surface than satisfied with current performance. Twenty-one answers is few; the report says so and so do we. But the grade isn't what interests us. The next question is: what do they measure it on?
The report answers. Asked which metrics matter when judging whether AI works, 69% cite operational efficiency, 48% lower CPMs, 45% faster or more accurate reporting. 17% have no formal measurement. The number one use of AI, cited by 22 respondents out of 29, is reporting and dashboards; measurement and verification only reach 34%. And asked to rank seven concerns, measurement, attribution loss in AI-mediated journeys, comes sixth, just ahead of environmental impact.
In other words, the market measures AI by the time it saves on tasks that already existed, and doesn't worry about whether it measures right. Reports produced faster, bids adjusted more often, a lower CPM. That's useful, and it's exactly what the platforms have been selling for three years. But none of those indicators says whether the campaign earned more than it cost. A lower CPM on useless impressions is still money lost.
We know it from experience: an agency that produces its reporting in two hours instead of two days has improved its margin, not your result. And 2.76 out of 5 is an honest grade for that: AI has made operations faster, not decisions better.
The missing measure is an old one; it's called incrementality. A MER, total revenue over total media spend, read every week. A geo test whenever a question arises: would this Advantage+ campaign have sold without us? These are measures of outcome, not productivity. They say what the agent actually contributed, and they also say when to stop it.
That is what we judge our agents on at mediagentic, and nothing else. Not hours saved, and for good reason: we don't bill hours. Not reports produced: they're produced continuously, and that no longer has value in itself since it has become a commodity. The written plan sets an incremental target, agents operate to reach it, a human reads the MER and decides.
The day the market grades its AI on MER rather than operational efficiency, the grade will change and become a true indicator of value created.