For most of our history, Intercom was a traditional SaaS business. Like all of our peers, we charged per seat. Our sales teams optimized around seat expansion, and we built entire systems (product, sales, incentives, forecasting) around the assumption that software value scaled with human usage.
AI broke that assumption.
When we set out to re-found Intercom as an AI-first company in 2023, we were solving complex technical problems while facing an equally consequential commercial challenge. We had to answer a deceptively complex question: In an AI-driven world, what are customers actually paying for?
What Is Outcome-Based Pricing?
Outcome-based pricing is a billing model where customers pay only for successful results rather than access, seats or usage. In this model, software companies shift from charging for tools to charging for accountability, earning revenue only when the product, such as an AI agent, successfully delivers a specific, predefined outcome for the user.
Pricing the Outcome, Not the Tool
When we launched our AI customer service agent, Fin, we made a decision that felt uncomfortable at the time: customers wouldn’t pay for access, seats or usage. They would only pay when Fin successfully delivered an outcome. If the customer wanted our AI agent to resolve the conversation, but it needed to be escalated to a human, customers wouldn’t pay. This was a major bet on accountability.
Seat-based pricing is forgiving because it allows software companies to get paid even when the value is unclear or aspirational. Outcome-based pricing does the opposite, forcing you to put your revenue directly behind your product’s ability to deliver results.
Internally, that changed everything, and outcomes became the company’s North Star. Product decisions, model improvements, customer onboarding and even sales conversations all started pointing in the same direction. There was no hiding behind feature roadmaps or promised ROI, which drove us to ship improvements faster because every failure was measured in lost revenue, not just churn risk.
While we were early to the game in adopting the pricing model, this customer alignment is why outcome-based pricing is spreading. Once customers experience software vendors sharing real risk, it’s hard to go back.
We Need a New Go-to-Market Model
Here’s what we learned quickly: When you only get paid if the AI works, you can't treat it like traditional software. Traditional SaaS deployments assume value emerges after setup, but AI systems need the right context, data and continuous feedback loops to perform as customers’ businesses evolve. We had to work closely with customers from day one, not to provide white-glove consulting, but to make sure Fin was configured correctly and improving continuously. Our pricing model made this essential, not optional. If we let customers struggle after setup, our revenue suffers immediately.
Outcome-based models reward speed, ownership and the ability to ship improvements that directly impact performance. So, it also changed how we build our teams. We prioritized hiring product-minded people who could identify problems, talk to customers, ship solutions end-to-end, and then evaluate the commercial impact of that change. Our core AI team is intentionally small and specialized, consisting of researchers and ML scientists who’ve shipped systems at scale. This combination lets us move fast without breaking trust because every improvement we ship directly translates into value for our customers.
Stabilizing Pricing Volatility
Outcome-based pricing is powerful, but it introduces bill volatility. Initially, customers couldn’t predict monthly costs, which created anxiety even when the AI was performing well. The big lesson we learned was the need to invest in ways to help customers adapt to the variable nature of the pricing. If we were starting again, we’d build these mechanisms into our holistic pricing and billing experience from day one rather than layering them in after launch.
The solution wasn’t to abandon outcome-based pricing, but to give customers tools to manage the variability. Take Rocket Money: They handle more than 60,000 support conversations a month, with Fin resolving 68 percent of them at $0.99 per resolution. A product launch or a seasonal surge can push that volume up sharply. Suddenly, customers are looking at a bill that moves in ways a traditional software budget isn’t designed to absorb. That’s not because the AI is failing, but simply because they couldn’t predict how well it would perform.
We built three mechanisms to solve this. Annual outcome buckets let customers purchase a block of resolutions upfront and draw them down over the year, so instead of Rocket Money facing an unexpectedly high bill in a peak month, that volume is absorbed against what they’ve already contracted for. Discounted overages mean that when customers exceed their contracted volume, they stay at their negotiated rate rather than being penalized for underestimating Fin’s performance. Product swaps address a third, connected problem: Companies moving from traditional models often carry seat-based contracts they no longer need at the same scale, and risk over-investing in human capacity before they understand what Fin can handle. Rather than waiting for renewal to rebalance, customers can redirect unused seat spend toward outcome credits mid-contract.
Together, these mechanisms didn’t compromise our outcome-based model; they actually made it scalable.
AI doesn’t just change what software can do. It changes how value is created, measured and shared. Outcome-based pricing is one expression of that shift, but it only works when companies are willing to rebuild around it. It introduces volatility into revenue, forces uncomfortable transparency with customers and exposes product weaknesses immediately. Outcome-based pricing is powerful, but it doesn’t come free. The founders who get this right won’t just price differently. They’ll build different companies.
