Enterprise DNA
News AI News

Thrive Capital's Kushner discloses real agent-ops numbers, then pushes back on AI hype

Josh Kushner's first-ever formal Thrive investor letter reveals concrete metrics from portfolio companies run with a 35-person engineering team: tax.

Enterprise DNA |
Thrive Capital's Kushner discloses real agent-ops numbers, then pushes back on AI hype

AI Pulse · Business Models & Winners

The play

Identify repeatable workflows with measurable accuracy and resolution rates, funding focused pilots instead of scattered AI experiments.

Josh Kushner’s first formal investor letter to Thrive Capital backers just gave us something rare in AI investing: real numbers from real companies, not projections. At Thrive’s accounting business, AI agents are getting tax returns done 30% faster while holding 98% accuracy. At its IT services business, agents are resolving half of all support tickets end to end, no human handoff needed. This is coming from a 35-person engineering team, which tells you the results aren’t buried in some massive R&D budget. They’re coming from a small, focused build.

That’s the part worth sitting with. If a lean team can get agents handling half of IT tickets on their own, most owners running support desks, back offices, or accounting teams have room to do the same without hiring an army of engineers first.

The second part of the letter is just as telling. Kushner is pushing back on the spray-and-pray approach to AI investing that’s dominated the last couple of years. About 90% of each Thrive fund sits in its 15 largest bets. That’s concentrated conviction, not a scattershot portfolio hoping something sticks. Coming from one of OpenAI’s biggest outside backers, that’s a signal worth noticing. The smart money isn’t chasing every shiny AI startup anymore. It’s doubling down on fewer, better bets and demanding proof they work.

For business owners, the takeaway is simple. The bar for “does this AI thing actually work” just went up, and so did the appetite for hard numbers over hype. If you’re building or buying agent-based tools, ask for metrics like these, not promises. This is the kind of operational proof point we like to track and build into an AI command centre, so the wins show up as real numbers, not slides. More detail is in the original report.

More like this?

Free daily email

Get this every morning.

This brief is one item from today's AI Pulse, the short daily read we run for ourselves on what is actually happening in AI. Subscribe free and it lands in your inbox each morning.

Free daily email

Subscribe to the daily AI Pulse

One short read every morning on what is actually happening in AI. Free.

One email a day. Unsubscribe any time.

Working With Claude field guide cover

Free Resource

Put what you just read to work

The free 32-page Working With Claude guide: the full ecosystem, Claude Code, and how to roll it out across a business.

Add your name (optional)

No spam. Unsubscribe any time.