What Launched

Kokai Zuma is built around Koa Assistant, a conversational AI layer backed by a growing set of specialized agents that handle audience building, campaign troubleshooting, and performance analysis on their own. The interface itself changed too, the dense "periodic table" view that used to be the default is now optional, replaced by a simpler, more guided workflow built around briefing the assistant rather than configuring every setting by hand.

Some of the deeper agentic capabilities are still in closed beta, but most of Zuma's features are already live for every advertiser on the platform, this is not a distant preview, it's shipping now.

This Is Bigger Than One Platform

Kokai Zuma is the latest and most explicit step in a trend that's been building across the industry: Google's Performance Max, Meta's Advantage+, and now The Trade Desk are all converging on the same idea, instead of a human setting bids, audiences, and creative rotations, an agent makes those calls continuously, in real time, based on outcomes rather than fixed rules.

For marketing teams, that changes the actual job. Less time gets spent adjusting settings inside a dashboard; more time gets spent writing the brief the agent works from, defining the guardrails it operates inside, and checking that what it's optimizing for actually matches the business outcome you want.

Read the Fine Print

A 32% average CPA improvement is an early, self-reported figure from initial results, treat it as a promising signal, not a guarantee for your account. Agentic campaign tools still optimize toward whatever signal you feed them, so weak conversion tracking or unclear success metrics will get automated and scaled just as fast as a good setup would.

3 Things to Do Before You Turn On Autopilot

  • Clean up your first-party data first. An agentic system optimizes against whatever signal it receives, inaccurate conversion tracking or thin audience data gets amplified, not corrected, once you hand over control.
  • Set explicit guardrails, not just goals. Define budget caps, brand-safety boundaries, and audiences that are off-limits before the agent starts making autonomous decisions, not after you notice something went wrong.
  • Keep a human reviewing outcomes weekly. Agentic tools reduce manual configuration work, not oversight. Someone on your team still needs to confirm the agent's definition of success matches yours.

Agentic advertising is moving from experiment to default faster than most marketing teams have adjusted their workflows for. The businesses that benefit will be the ones that treat these agents as a powerful new hire that still needs a clear brief and regular check-ins, not a switch you flip and walk away from.