When the major agency groups talk up their AI investments, they’ll emphasise the new possibilities around personalisation and creativity at scale which the technology opens up. But at the moment, a new survey from IAB Europe suggests that pure operational efficiency remains the most common metric through which businesses across the industry are judging their AI investments.
The path to full autonomy
IAB Europe’s 2026 Impact of AI on Digital Advertising Report surveyed 50 respondents from agencies, ad tech companies, advertisers, publishers and other organisations to gauge how companies across the advertising landscape are adopting and adapting to AI.
Unsurprisingly, total adoption is high. Ninety-two percent of those surveyed said their businesses already use AI tools internally, while 86 percent reported using AI specifically in marketing processes.
Reporting is the most common AI-supported workflow, cited by 76 percent of the respondents who gave detail of where AI is being deployed. Programmatic optimisation was the next most frequent use case, with 59 percent, followed by media planning and strategy (55 percent), content creation and source tracking (48 percent), and dynamic creative optimisation (48 percent).
Agentic AI implementations still generally require a fair amount of human input. IAB Europe defined five levels of agent autonomy, from agents acting based on users’ directions and decisions (level one), to agents operating autonomously under monitoring (level five). The majority of respondents said they either have no agentic systems in place, or they use agentic systems which people direct or plan with. It’s notable though that a couple of companies claimed to sit at level five, with full agent autonomy.
Short-term struggles
There are mixed beliefs around how quickly agentic systems will grow to be a major part of day-to-day workflows. Thirty percent of those surveyed said they expect agentic ad buying to become a main trading method in some markets over the coming year. At the other end of the spectrum, eight percent don’t expect any value from their AI investments in that time frame.
Indeed, there seems to be a general tension between how AI is currently delivering, and what’s expected of it one year from now. When asked to rate AI’s performance in ad operations from one to five, responses averaged out at 2.76, which seems decidedly mediocre. And for now, the metrics AI is being measured against are fairly basic.
Sixty-nine percent of those who answered the question said that operational efficiency is used to evaluate AI tools. CPM improvement (48 percent) was the second most common answer, followed by reporting accuracy or speed improvement (45 percent).
Perhaps this tension isn’t surprising, though, given the rate of AI development. Embedding new systems and understanding what they’re fully capable of takes time, and basic workflow efficiencies are some of the easiest, earliest wins to extract from AI. But given the rate at which agentic capabilities are evolving, it’s understandable if there’s a significant gap in what AI is delivering today, and what it’s expected to do in one year’s time.
Follow VideoWeek on LinkedIn.
