At Cannes, WPP put Google DeepMind's AlphaEvolve to work auto-proposing and refining the models behind its campaigns, lifting prediction accuracy ~10%, and used Gemini to relip the Duracell Bunny across languages for a single global spot.
Why it matters: When the optimization layer itself goes agentic, the durable edge moves up to strategy, briefing, and measurement. The parts an agent can't own.
Agentforce Commerce's Shopper, Buyer, and Merchant agents are now GA with native ChatGPT, Gemini, and Google AI Mode hooks, and Salesforce says retailers already running shopper agents grew sales 59% faster than those still on the sidelines.
Why it matters: Commerce is moving onto the AI-answer surfaces buyers now start on. Being cited and agent-readable there is becoming the new top of funnel.
The deal folds Aampe's per-user agents, 200B+ decisions a week across brands like Grab and Swiggy, into MoEngage, collapsing campaign workflow and 1:1 personalization into one system where marketers set goals and guardrails, not individual sends.
Why it matters: Lifecycle is consolidating around autonomous per-user agents. The work shifts toward defining objectives and guardrails rather than pushing buttons.
Forrester's State of AI in US Marketing Agencies (with the 4As) finds 9 in 10 use genAI and half use agentic AI for execution, but warns the focus on efficiency (74% just summarizing docs) is eroding creativity and brand growth.
Why it matters: The gap between using AI and using it to actually grow is exactly where sharper operators win the work.
Magnite, Mediaocean, and MiQ are shipping agents that turn a brief into a live, self-pacing campaign across platforms, pushing autonomy out of creative and into the media-buying seat itself.
Why it matters: As buying automates, differentiation moves up to positioning, creative, and the first-party data feeding the agents.