Paul Meinshausen describes marketers’ belief in control as a “control illusion.” He says leading AI agents resembles managing new team members: set goals, provide training and governance, then let them experiment. If marketers dictate every action, they defeat the agent’s scaling purpose.
Meinshausen says personalization has mostly enabled humans to run more complicated rules, not true one-to-one experiences. Those rules hit a complexity ceiling and do not scale well. From the customer’s view, the result still does not feel personal.
Paul Meinshausen contrasts product search, where customers pull what they want, with marketing, where teams must push proactive messages into an attention-scarce world. Marketing gets less data and faces a harder task: guessing what someone might want before they ask.
Meinshausen says brand consistency is more illusion than achieved standard, with campaigns crossing ops, data and design teams. He argues agents can pull from Figma and generate on-brand versions more fluidly. Leaders should aim to get much better, not merely avoid getting worse.
Paul Meinshausen says marketing still measures aggregate averages for a campaign sent to a million people. Agentic personalization needs counterfactuals and incrementality at the individual user and message level. He compares this to Netflix measuring movies at scene level rather than asking if viewers liked the film.
Meinshausen explains AMP built an agent for every end user but still needed CRM and customer engagement systems for delivery. Rather than replace complex stacks, agentic tools must fit through interoperability and modularity. That overlap led AMP to join MoEngage rather than build every adjacent layer alone.
