Come and Talk to us

The distance between a working AI pilot and something running in production is almost never about the model. It is about who owns the output when it is wrong, which team absorbs the extra review load, and what gets decommissioned to make room. Those are organisational questions with organisational costs, and they are the ones nobody puts on a conference slide. Three days is enough time to have that conversation properly, which is why we are bringing a stand rather than just a delegate pass.

Masters is the densest gathering of senior client-side marketers on the calendar and the worst place on earth to have an unplanned conversation. Everyone is between sessions, everyone is being approached, and nobody has twenty free minutes by accident. So we run this one differently: a stand for people who want to drop by, and a diary that is mostly booked before anyone gets on a plane. If you want time with us here, lock it early rather than hoping to find us.

Everyone agrees the answer is AI plus human judgement. Almost nobody can say where the line sits. The useful version of this conversation is specific: which decisions you automate, which you refuse to, and how you can tell the difference before a campaign ships rather than in the post-mortem.

Most of what is currently called AI visibility measurement is SEO thinking with the nouns swapped. Rank, share of voice, impressions. None of it maps cleanly onto a system that synthesises one answer rather than ranking ten links. We are attending to see which approaches survive scrutiny, and we will write up the ones that do.

Fragmented measurement usually gets blamed on tooling, so the response is another tool. The failure is upstream of any model: nobody named the decision the measurement was meant to inform, so three methods produce three answers, none of them wrong and none of them used. Bring us a dashboard nobody acts on and we will tell you which question is missing from it.

Retail media reports on its own metrics. CTV reports on different ones. Incrementality testing runs on a third basis entirely. Each is defensible alone, and the moment you put them side by side to make a budget decision the comparison is guesswork wearing a chart. We think that is the real measurement problem this year, and it is far less interesting to talk about than AI, which is exactly why it stays unsolved.

Attribution across an eighteen month cycle, through a distributor network you do not control, to a buying group you cannot fully see. Nobody has a clean answer and most vendors pretend otherwise. We are attending because the honest attempts are worth more than the polished ones, and we will write up whichever we hear.