Outcome-Based Pricing Solves AI's Trust Problem. It Creates a New Budgeting One.

When technology works perfectly, it disappears. AI customer service is the cleanest example of this, the best resolutions are the ones nobody mentions, because the customer got an answer and moved on without ever feeling friction.

That's also exactly why it's hard to charge for. Invisible infrastructure is notoriously difficult to monetize, because the person renewing the contract forgets what it's quietly doing for them. A support team that used to handle 5,000 tickets a month and now handles 2,500 doesn't generate a headline. It generates a budget line that looks suspiciously easy to cut.

Intercom's pricing for Fin, its AI customer service agent, is a direct answer to that problem. Fin is priced at $0.99 per outcome, a resolved conversation or a completed procedure handoff, and you're billed once per conversation no matter how many questions Fin actually answered along the way.

That's a positioning decision dressed up as a billing mechanic. Every $0.99 line item becomes a micro-receipt: proof, attached directly to the invoice, that the infrastructure did something specific. Buyers don't have to take a vague "efficiency gain" on faith at renewal time. They can count the receipts.

Outcome-based AI pricing charges customers only when the AI delivers a defined result, a resolved ticket, a booked meeting, a qualified lead, instead of for seats, tokens, or access to the software. Fin is the clearest live example: $0.99 per resolved conversation, billed once regardless of how many steps it took to get there. The model turns an invisible efficiency gain into a line item buyers can point to, which is precisely the kind of proof traditional SaaS subscriptions have always struggled to produce for AI features.

How "resolved" gets defined matters more than the headline price. Intercom counts a resolution when the customer confirms Fin's answer helped, or simply leaves the conversation without asking for more, and it doesn't bill you for conversations where Fin detects frustration and proactively hands off to a human. That second part is a quiet but important design choice. It removes the incentive to count a frustrated customer's silence as a win.

It's worth being honest about what "it works" means in practice, too. Intercom showcases customers like Synthesia, where resolution time reportedly dropped by roughly 96% after deploying Fin, a strong number, but a marketed one. Independent analysis of Intercom's public case studies puts real-world resolution rates in a steadier 42–50% range, which is the number worth budgeting against, not the best-case headline.

Bessemer Venture Partners' framing of the model is the sharper version of the same point: $0.99 per resolution doesn't just price the product, it aligns sales, support, and product around one shared outcome, resolved tickets, instead of three different definitions of "engagement." That's the real shift. Pricing isn't just a finance decision anymore. It's an organizing principle.

Not every vendor calling itself "outcome-based" actually is one. Salesforce's Agentforce charges $2.00 per conversation regardless of whether the issue gets resolved, if the AI fails and a human steps in, you still pay. That's usage pricing wearing outcome language. Intercom's "you only pay when it works" claim is the actual test of a genuine outcome model: does the vendor get paid on your failure, or only on your success?

The Catch Nobody Puts on the Pricing Page

Here's the trade buyers are making, whether they've noticed it or not: in a true outcome model, your bill scales with your own success.

Run a demand gen campaign that spikes product usage, and more customers hit your support flow. Fin handles more conversations. Your monthly AI bill goes up, not because Intercom changed anything, but because your marketing team did its job. Most CFOs don't experience that as alignment. They experience it as a forecast they can't trust.

This isn't theoretical. Reported jumps in Fin bills include swings from $4,000 to $9,000 a month, and from $1,200 to a projected $10,000, as resolution volume and rate moved. Per-seat pricing was annoying, but it was legible. You always knew your headcount. Resolution pricing breaks that legibility, because it scales with how much your customers use the AI, not with anything procurement directly controls.

It compounds at scale in a specific way, too. At high conversation volumes, the Fin line item alone typically drives 70–80% of the total bill, not the underlying Intercom seat cost. The better your resolution rate gets, the more of your invoice depends on a number you don't set.

Play with the resolution-rate slider for thirty seconds and the problem becomes obvious. The same volume of tickets can produce a wildly different bill depending on a number, AI resolution rate, that the vendor controls more than you do.

What This Means for How You Actually Buy It

None of this is an argument against outcome-based pricing. It's an argument against buying it pure.

The fix is already showing up in the market. Bessemer's research found that roughly 92% of AI software companies now use some form of mixed pricing with a usage component, and hybrid models, a base subscription that covers fixed costs and a revenue floor, paired with an outcome-based charge above that baseline, are projected to cover 61% of SaaS companies by the end of 2026. Gartner expects at least 40% of enterprise SaaS spend to have shifted to usage-, agent-, or outcome-based pricing by 2030. The direction is settled. The pure version of any of these models isn't.

A hybrid structure does the thing pure outcome pricing can't: it caps the vendor's exposure to AI failures, and it caps the buyer's exposure to AI success. Both sides get a floor.

If you're evaluating a per-outcome AI agent, model the bill at your highest plausible volume, not your average one, and ask the vendor directly what happens to your per-unit cost as resolution rate improves. That's usually where the bill quietly stops being about seats and starts being about something you have far less control over.

Intercom is right about the thing that matters most: AI value should show up as proof, not promise. The next move, for Intercom, and for every vendor following the same playbook, is pricing that proves the outcome without making the buyer's January budget a guess.