One of the biggest mistakes companies make about churn is waiting to snap into action only once a customer requests to cancel. It starts months earlier, in how they were acquired, how they were onboarded, and whether they are getting value from the product and pricing model, giving them reason to stay.
I spent years building a company around the cancellation moment. What I learned: by the time a customer reaches it, you’re reading the autopsy of decisions that were already made.
I hear the same thing from founder after founder: customers often churn before they’ve ever reached a moment of real value. Not because the product failed them. Because the path to value was never built.
The fix lives much earlier than most teams are looking.
The first 90 days determine lifetime value
Most AI products don’t deliver full value out of the box. They need configuration, training, and context before they work reliably for each customer’s specific situation. That’s not a flaw. It’s the nature of the technology. But it means the onboarding burden is real, and customers left to figure it out alone often won’t get there. They’ll hit a frustrating early experience, decide the product doesn’t work, and cancel without ever seeing what it can do.
Kunal Agarwal, CFO of Gorgias, described this at our Beelieve event in San Francisco. Gorgias processes over 300 million customer conversations and saw 350% growth in AI agent usage over the past year. When they analysed their churn data, one number clarified everything: customers using the platform at 70% of capacity or above were dramatically less likely to cancel. That figure became the target their entire onboarding investment was designed to hit, not a metric tracked after the fact, but the working definition of a successfully onboarded customer.
The investment that got customers there was human-led, or more specifically forward-deployed engineering: configuring the agent to each merchant’s specific workflows, guiding setup, and staying present through the first weeks of live deployment. On paper it looks expensive. The calculation changes when you measure what that retention difference means to lifetime revenue.
As Kunal put it: “It may seem like, wow, this is really expensive to get this customer onboarded, but it’s worth it if you think about the difference between the proclivity to churn versus not.”
The first 90 days of an AI subscription are more predictive of long-term customer value than the first year of a traditional software contract. Much of what teams attribute to product quality is an onboarding failure. That’s actually good news. It’s a solvable problem.
But getting customers to 70% only solves half the problem. The other half is whether you sold them the right amount to begin with.
Pricing that aligns behaviour and value drives growth and retention
Pricing is where I see the most avoidable damage. Not because teams price incorrectly, but because they see their first pricing model as a hardened strategy (backed by a revenue forecast), versus as an exploratory approach needed to validate the adoption process and usage assumptions. In a market where customers are still forming habits around AI products, pricing complexity introduced before adoption is established creates friction that can lead to churn and cancellations, but is actually something simpler and often avoidable.
Gorgias ran into this directly when it launched outcome-based pricing: customers pay for fully resolved AI interactions, not raw usage. That was the right structural call. The next decision was harder: should they charge more for sales-assist interactions than for support resolutions? An AI agent that helps convert a $250 shoe purchase creates different value than one that automates a $5 support ticket. The logic was sound. But when they tested it, customers pushed back, not on the price, but on the complexity. They were still figuring out how to build AI-native workflows into their companies’ way of working. Introducing a two-tier model at that moment raised awkward questions about whether AI or humans in the loop should get the credit for the outcome, slowing adoption rather than capturing more value.
Gorgias simplified it to a flat fee per resolved interaction. “Are we leaving money on the table? Absolutely,” Kunal said. “But we’re optimising for adoption.” The plan to revisit differentiated pricing for sales outcomes came after that simplification earned them a customer base with durable habits, not before. That sequencing is the lesson. Earn the mental model first. Capture differentiated value second.
The 70% usage threshold is also a selling instruction. A customer contracted above what they’ll realistically consume starts below that threshold before they’ve had a fair chance to reach it. Undersell and create a clear path to expand. A customer who grows into their contract and upgrades delivers more lifetime value than one who overpays, under-uses, and leaves convinced the product underdelivered, even when it didn’t.
The opportunity is in the sequencing
Churn in an AI product is a continuous briefing on your market. It tells you whether your acquisition motion is pulling in the right customers, whether your onboarding is getting them to value before they get distracted or give up, and whether your pricing is asking them to understand complexity before they’ve experienced any benefit. In AI, that lag is more costly than it’s ever been, because the feedback loop is faster and the early churn window is shorter.
Getting the sequencing right is a decision you make before the churn data tells you to. The AI founders I speak with have arrived at the same conclusion by different routes: acquisition, onboarding, pricing, and retaining aren’t afterthoughts to the product. They’re part of it. The ones who are accelerating from strong customer retention and growth have focused on defining product-market fit not only by how well they acquire and sell customers, but by their time to go live, achieve measurable success, and willingly pay more for more AI usage – with the same rigour they applied to building the product itself.
Guy Marion
Guy Marion is Chief Marketing & Growth Officer at Chargebee. A founder and revenue operator with nearly 18 years of experience, Guy has grown software businesses from $0 to $175M+ ARR across sales, product, marketing and growth.
