Closed-won lookalikes beat another agent demo

Multi-agent GTM theater without closed-won patterns is just software with a better demo reel.

On Oct 1, 2026, ZoomInfo announced Agent Teams in GTM Studio, built on the orchestration architecture of DoubleO.ai (which it acquired the day before), including champion tracking and closed-won lookalike plays. That launch is a product story. The GTM lesson is older than the press release.

The durable asset is the lookalike of what already closed, plus who can still champion it.

Why do closed-won lookalikes matter more than the agent brand?

Because about 95% of B2B buyers are out of market at any given time (Ehrenberg-Bass Institute research for the LinkedIn B2B Institute). You cannot afford to aim at "everyone who looks like a logo."

A closed-won lookalike starts from deals that already paid. Industry, size, stack, trigger, and the job the product actually did. That is a pattern you can defend in a Monday pipeline review. An agent that can draft, research, and sequence without that pattern is still guessing at volume.

I do not care which vendor put "Agent Teams" on the blog. I care whether your team can answer: which accounts look like last quarter's wins, and which people inside them can still sponsor a deal.

What is a champion signal in practice?

A live map of who pushed the last closed deal, who still works there, and who replaced them when they left.

Champion tracking is not a LinkedIn scrape of "decision makers." It is CRM truth plus change detection. Who was on the mutual plan. Who replied first. Who brought budget. Who left. Who is still warm enough to reopen a thread.

That is why warm account vs warm contact matters. An account can look hot while the champion who made it hot is gone. Lookalike without champion quality is still a cold list with nicer filters.

How do lookalike plays fit signal-based outbound?

They sit upstream of the note. Downstream of the win.

Signal-based outbound needs a reason to believe the account is in a buying window. Closed-won lookalikes narrow who is worth watching. Then you layer live signals:

1. Site de-anon on pricing and compare pages. Company resolution on your own domain. I covered the mechanics in website de-anonymization for B2B GTM.
2. G2 category and competitor traffic. Review and compare behavior where you sell. That is how G2 category pages become buying signals.
3. Champion change. New VP, new ops lead, old champion at a new company that still matches the win pattern.

The play is not "run more agents." The play is score accounts that look like closed-won, confirm a champion still exists, then act when a live signal shows up.

What do most teams get wrong after a launch like this?

They buy the theater and skip the pattern work.

They turn on multi-agent research, draft, and send. They do not export closed-won with firmographics, use case, and champion names. They do not kill lookalike segments that never book meetings. They treat the product name as the strategy.

If you cannot describe your last ten wins in one table, an agent team will just scale the mess. Lookalike quality is a CRM and messaging problem before it is a tooling problem.

What should I do this week?

Build the win table before you expand the agent stack.

1. Export closed-won from the last two quarters. Company size, industry, primary use case, ACV band, sales cycle length, champion title, lost reasons on near-misses.
2. Name three lookalike segments you would actually call. Kill anything that cannot earn a Monday review.
3. Map champions on open opps and recent wins. Flag titles that churned. Rebuild the intro path where the champion left.
4. Wire one live signal per segment. Site visit, G2 compare, or product usage. No signal, no send.
5. Score meetings by lookalike segment for 30 days. If a segment never books, delete it. Do not feed it to a smarter agent.

Adapt or fail. Pattern first. Theater second.

Start Signals, Convert, Grow

FAQ

Did ZoomInfo really ship closed-won lookalike plays?

ZoomInfo's Oct 1, 2026 announcement on Business Wire says Agent Teams in GTM Studio include champion tracking and closed-won lookalike plays, built on DoubleO.ai's orchestration architecture. Check ZoomInfo's help docs for what is live in your account before you buy the story.

Is a lookalike the same as ICP?

No. ICP is who you say you sell to. A closed-won lookalike is who already bought, with the traits that showed up in the win. Start from wins. Then tighten ICP.

What if my CRM champion data is garbage?

Fix the data before you scale outbound. A bad champion field plus a lookalike engine is still a cold list. Require champion title and status on every closed-won and open opp for 30 days.

Can I use G2 and site de-anon without a lookalike model?

Yes, and you should. Lookalikes tell you who is worth watching. G2 and de-anon tell you who is moving now. Use both.

Does this replace intent data vendors?

No. It tells you whether the vendor's "similar accounts" match your actual wins. Your closed-won table is the scorecard. Their model is a suggestion.