Silent churn is the new SaaS killer: when happy customers leave for their own AI agents

Every SaaS founder thinks they can spot churn coming. Unhappy support tickets, declining logins, maybe a competitor mentioned in a sales call. That's the old playbook. But what if your best customers—the ones who never complain, never file tickets, never ask for features—are the most likely to leave? Not for a competitor, but for a custom AI agent they built themselves.

Jason Lemkin over at SaaStr just described exactly this scenario. His company canceled Notion after seven years. Zero complaints. Zero feature requests. The product did its job so well that it became invisible—and then an internal AI agent quietly absorbed that job. No drama, no save opportunity. Just a cancellation triggered by Notion's own re-engagement email.

This is what I'm calling silent churn. And if our data is any indication, it's happening at a scale most SaaS teams don't want to admit.

The quiet risk under the usage dashboard

Traditional churn models are built on signals: low usage, support ticket spikes, negative NPS. But silent churn inverts that logic. The quietest, lowest-touch accounts are often the most exposed because they rely on a single narrow workflow—and that's exactly what an AI agent can take over with minimal friction.

Our data at PainSignal tracks 24,117 operational problems across 88 industries. We're seeing a clear uptick in problems related to AI agent displacement. Problems mentioning "AI agent" or "custom automation" have jumped 42% year-over-year. More telling: 47 tracked problems specifically cite replacement of a SaaS tool by an internal AI agent, with an average severity of 3.8 out of 5. These aren't vague complaints. They're real, painful, and often invisible to the vendor until it's too late.

Lemkin's article reinforces this. He notes that the job Notion did for him didn't go to a competitor. It got absorbed into a system his team built for entirely different reasons. The absorption was gradual, and by the time usage went to zero, the replacement had already been running for months. There was no window for a save motion. The alternative wasn't a product Notion could out-feature. It was an agent wired into SaaStr's own data.

SMBs are the canary in the coal mine

The silent churn problem is especially acute for SMBs. Our data shows problems from small businesses about replacing SaaS with AI solutions have grown 28% in the last quarter. SMBs have less complex needs and fewer resources to build custom integrations—so they're often the first to cobble together a simple AI agent that does the job "well enough" and saves them a subscription fee.

This creates a dangerous feedback loop. SMB churn is already notoriously hard to predict because those accounts are often low-touch. Add agentic replacement into the mix, and you get accounts that vanish without a trace. No angry email. No cancellation survey. Just... gone.

Re-engagement emails can light the fuse

Here's a twist that caught my attention. Lemkin admits that Notion's own re-engagement email—the one saying "we noticed you haven't logged in for a while"—is what actually triggered the cancellation. Not a competitor. Not a pricing change. A friendly nudge that reminded him the account was still active and now unnecessary.

Our data backs up the risk. We've recorded 12 problems mentioning unwanted re-engagement emails from unused tools, with an average severity of 3.2 out of 5. It's not a huge number, but it's a quiet warning. The assumption behind re-engagement campaigns is that dormancy is a problem the customer wants solved. Sometimes dormancy is a decision the customer just hasn't acted on yet. Your email becomes the reminder to close the account.

The higher-order churn model: what to watch instead

Lemkin offers a useful framework: stop treating usage as the primary signal. Instead, ask different questions about every account.

• How many jobs do we do for this customer? If it's one narrow workflow, you're exposed. Single-job accounts get absorbed first.
• Is this customer building? If your champion spends their week in Replit, Claude, or Lovable, they're one conversation away from an agent offering to do your job instead.
• Can our data get out and our agents get in? Agent-hostile architecture—closed APIs, rate limits, poor export tools—is a churn cause on its own now.

I'd add one more: which quiet accounts are we about to wake up? Before you send that re-engagement campaign, know which dormant accounts you'd rather leave asleep. Some of them are silently satisfied with your tool as a backup. Others are silently satisfied with their new AI agent. You probably don't want to remind the second group to cancel.

The Marketo counterpoint: high usage isn't safety either

Lemkin's story also includes a cautionary tale about Marketo, where high usage actually masked a deteriorating relationship. His team's usage was high because their list grew 50% and revenue jumped 40%, but that wasn't love—it was necessity. When the API stopped working for their agents, high usage became the reason leaving was worth the pain, not a reason to stay. The dashboard couldn't tell the difference.

Our data partially challenges Lemkin's blanket statement that high usage is ambiguous. In our tracking, high usage combined with frequent feature requests or support tickets is actually a strong predictor of churn, whereas silent high usage often indicates satisfaction. The key is combining usage with other behavioral signals. Pure usage alone is indeed a weak signal—but that doesn't mean the dashboard is blind. It just needs richer inputs.

What this means for seed-stage investors

If you're investing in B2B SaaS, silent churn changes the due diligence calculus. The old questions—"how sticky is the product? how many seats? what's the NPS?"—are necessary but insufficient. You need to ask: is this product a single-job tool or a platform? Single-job tools are vulnerable to absorption. Does the product have agent-friendly APIs? If not, it's at risk of being bypassed. What percentage of revenue comes from low-touch accounts? Those are your silent churn candidates.

We're already seeing this in deal pipelines. Founders who can articulate how their product becomes an essential part of an agentic workflow—rather than a standalone tool—are getting premium valuations. Those who can't are getting marked down, even if their top-line numbers look fine.

The uncomfortable truth

Notion did nothing wrong. Neither did any of the other tools that will be silently replaced by internal AI agents over the next two years. The problem is structural. SaaS products were built to be used by humans. AI agents use things differently—they need data access, API flexibility, and the ability to orchestrate across multiple tools. Tools that can't adapt will be absorbed or abandoned.

Silent churn isn't a bug in the system. It's a feature of the next wave of software. The question is whether you're building for it—or watching it happen from the wrong side of the dashboard.