Below this price, AI products churn 3x worse. Most founders price below it.
What 3,500 software companies reveal about why the same AI product can compound like SaaS or churn like a free trial and why it isn't the product.
👋 Hey, Sahil here - welcome to this edition of Venture Curator, where we break down how great startups grow, how top investors think, and what’s shaping the future of tech.
P.S. Get access to 100+ startup & VC resources, investor databases, fundraising templates, 150+ premium archive posts and exclusive startup research - all in one place.
FROM OUR PARTNER - PLANE
Webinar: How I Used AI to Automate Sales Ops and Scale to Multi-Million ARR
Scaling sales ops usually means more headcount. Plane co-founder Staszek Kolarzowski (YC ‘17) took the opposite route: he built an AI engine to run his, on the way to multi-million ARR. In this founder-to-founder webinar, he’ll give live demos of what’s in production today and what you can build yourself. You’ll walk away knowing how to:
Run growth experiments that track campaign performance and flag issues before they hit results
Deploy agent-powered sales coaching that trains reps and tracks skill development
Monitor pipeline health, auto-build win plans, and surface deal gaps before they stall
Calculate commissions accurately and sync them straight to payroll
Register for the free webinar →
📜 DEEP DIVE
Below this price, AI products churn 3x worse. Most founders price below it.
There’s a number quietly haunting AI founders this year, and it isn’t ARR.
It’s the share of last month’s revenue that’s still here this month. Unglamorous, hard to fit in a launch tweet and the cleanest signal of whether what you’re building compounds or quietly bleeds out.
We spent two years cheering the zero-to-$100M-in-a-year stories. They’re real. The part that rarely gets said out loud: that growth gets nearly impossible to hold once a bigger and bigger install base walks out the door every month. There’s a blunt name for it - burning through your TAM. You’re not growing; you’re refilling a bucket faster than it drains and calling the inflow “growth.”
So someone went and got the data - roughly 3,500 software companies, sorted into B2B SaaS, B2C SaaS, and AI-native, and compared on how well each held onto revenue. The first cut looks damning.
B2B SaaS: a median net revenue retention of 82%. Healthy.
B2C: 49% - leakier, as you’d expect.
AI-native: 48% net, but just 40% gross - worse than consumer apps.
AI was shedding customers faster than B2C, with no upsell to cover the loss.
If you stopped there, you’d conclude AI products are just structurally bad at retention. A lot of people stopped there. Cassie Young at Primary Venture Partners has been warning about an incoming “gross retention apocalypse,” and the chart above is the kind of thing she’s pointing at.
But the average is hiding the actual story. Because when you split those same AI companies by price, the “AI churn problem” doesn’t hold up as one problem. It splits cleanly into two, and which one you have was mostly decided by a pricing-page decision you made before launch.
Here’s the split:
AI products under $50/month: 23% gross retention, 32% net.
AI products at $50–249/month: 45% gross, 61% net.
AI products over $250/month: 70% gross, 85% net - essentially indistinguishable from healthy B2B SaaS.
Same underlying models. Same wrappers, half the time. What separates a 32%-net death spiral from an 85% compounding machine isn’t the quality of the AI - it’s the price in front of it.
Which flips the diagnosis. “AI has a churn problem” sends you off to rebuild the product. But this is a packaging problem, and packaging is something you decide before you ever ship.
That’s the good news and the uncomfortable news at once: you can fix churn without a heroic roadmap, but only if you can see what’s happening under your topline. Most founders can’t. They’re reading the wrong number, off the wrong month, and drawing exactly the wrong conclusion about who their customers are.
Here’s what this issue gets into:
Why is the retention number on your dashboard almost certainly overstating the truth, and which month should you actually be reading it from?
What’s quietly bleeding cheap, broadly-applicable AI products dry and why won’t a slicker onboarding flow stop it?
Where does the durable line really sit, if crossing it has almost nothing to do with the number on your pricing page?
Which three levers lift retention 10-20 points without shipping a single feature?
And what exactly should you run against your own numbers this week?
If you’re building anything AI-native or signing off on the budget for it, this is the part that decides whether the ARR you’re celebrating is the kind that stays.
Subscribe to Venture Curator Premium and get 30% off, forever.