--- title: "Is Outcome-Based Pricing Real, or Just Marketing Hype?" date: 2026-02-15 description: "Only 2% of companies used outcome-based pricing in mid-2025. By December it was 18%. The gap between the narrative and reality tells you something important about where AI pricing actually is." canonical: https://www.tansohq.com/blog/is-outcome-based-pricing-real author: Kat Laszlo --- Outcome-based pricing keeps coming up. In decks, on panels, in investor conversations. But almost nobody is actually doing it. Intercom Fin charges $0.99 per resolved conversation. Sierra gets paid when it completes a task. Zendesk announced a shift away from per-agent pricing last August. After that, the list gets thin. ICONIQ's Q2 2025 survey found only 2% of companies were using outcome-based pricing. By December, it was 18%. That's growth, but it's still a fraction. Most companies moving off seat-based pricing are going toward consumption models, not outcomes. And only 15% say they plan to move toward outcome-based pricing at all. So there's a real gap between how much people talk about this and how much is actually happening. --- ## First, a distinction most people skip Part of the reason this debate stays confused is that "outcome-based pricing" collapses two separate decisions into one phrase. One is the value metric: what do you charge for? Seats, tokens, API calls, resolved tickets, processed documents. The other is the pricing model: how do you collect the money? Subscription, prepaid credits, pay-as-you-go, arrears. When someone says "we do outcome-based pricing," they might mean they charge for resolutions (a value metric). Or they might mean they only get paid when the AI succeeds (a cash mechanism). Or both. Intercom charges per resolution on a pay-per-event basis. Salesforce charges per action via prepaid credits. Both get called "outcome-based pricing" even though they work completely differently in operations, cash flow, and risk. Once you pull these apart, most of the questions about outcome pricing turn into design choices. An outcome metric with a prepaid cash mechanism gives you value alignment without budget unpredictability. An activity metric with a success bonus gives you some outcome alignment without solving attribution. How you combine the two matters more than what you call it. Almost every problem people have with "outcome-based pricing" is really a problem with one half of this, not both. --- ## The idea is appealing. That's partly the problem. On paper, it works. The buyer only pays when the product delivers. The vendor takes on risk. ROI is straightforward to calculate. Investors like it because it signals confidence in the product. Shipping it is a different story. AI doesn't have a standard set of success metrics yet. In advertising, ROI is clear: a click, a conversion, a dollar. In AI, teams try to measure value through resolution rates, time saved, stickiness, retention. None of those have converged into an industry standard. Without a shared definition of success, you can't price on it. Attribution makes this harder. Who gets credit when an AI agent resolves a ticket but a human set up the workflow, curated the knowledge base, and handled the escalation? Partial outcomes, shared credit, counterfactual attribution. All unsolved. Siena AI's CEO wrote [a full post](https://www.siena.cx/blog/conversation-vs-outcome-based-pricing-ai-agent) arguing that outcome-based pricing actively hurts customers. The key observation from working with hundreds of companies: customers spend hours every week arguing with vendors about what counts as "resolved." Every vendor has their own definition. There's no standard. More than one founder has told me, unprompted, that outcome-based pricing is basically just usage billing with a different label. You're still metering events. You're still generating an invoice. You've just tagged some events as "outcomes." As one pricing strategist put it, outcome-based pricing is just "a marketing buzzword." --- ## Where it actually works It does work, but only in specific conditions. And when you look at the cases, the conditions are the same every time. | Condition | Required? | Why | |-----------|-----------|-----| | AI acts autonomously | Yes | If a human is in the loop, you can't attribute the outcome | | Completion signal is clear | Yes | "Resolved" needs a binary definition both sides agree on | | Cost variance is bounded | Yes | Vendor absorbs risk; unbounded cost kills margins | Customer service fits cleanly. Intercom Fin resolves a ticket, the resolution is binary, cost per resolution is roughly predictable. They saw 40% higher adoption versus seat-based pricing. That's real. Chargeback recovery, collections, specific legal workflows - same pattern. AI acts alone, the outcome is measurable, the vendor can pool risk across a customer base. It falls apart with copilots. Writing assistants. Advisory tools. Anything where a human is in the loop. If your AI is facilitating a human analyst rather than replacing one, there's no clean way to say the AI produced the outcome. The product architecture determines whether you can credibly meter outcomes - not whether you want to. And then there's the cost side that nobody in the boardroom talks about. If you price on a fixed fee per outcome, you're absorbing all the variance in what that outcome costs to produce. One resolution might take hundreds of tokens. Another might take millions. If you can't price differently based on how much work went into the outcome, you're running a slot machine. | Category | AI role | Can meter outcomes? | |----------|---------|---------------------| | Customer service | Autonomous agent | Yes (resolutions) | | Chargeback recovery | Autonomous agent | Yes ($ recovered) | | Legal automation | Autonomous agent | Yes (cases completed) | | Code completion | Copilot | No (who wrote it?) | | Writing assistance | Copilot | No (who wrote it?) | | Data analytics | Copilot | No (who queried it?) | --- ## Strategy meets the invoice Salesforce is the highest-profile case study. Three pricing iterations in 18 months. Flex Credits, per-action, per-user, hybrid. On earnings calls it sounds like deliberate evolution. On the operations side, the picture is different. The word I keep hearing from people close to Agentforce billing is "mess" - unclear how much is actually being transacted, hard to reconcile internally. Salesforce Ben's ecosystem report found the real roadblocks were unclear pricing, messy orgs, and weak enablement. A Monetizely analysis called the multi-model approach "design-by-committee in pricing." This isn't specific to Salesforce. Everyone building in this space runs into the same thing. If a company with that scale of engineering and go-to-market resources is struggling to operationalize flexible AI pricing, what does that look like for a 50-person startup trying to ship outcome-based billing with a Stripe integration and a spreadsheet? In practice, the bottleneck isn't choosing the right pricing model. It's confidence. Teams outline multiple hypotheses and it takes a month just to launch changes. They can't predict which ones will succeed. They want to know the outcome of a pricing experiment without actually running it. No tool solves that yet. A question I keep coming back to: do teams actually want to run pricing experiments, or do they just want the answers without running them? --- ## What the renewal cliff forces The SaaSpocalypse makes this urgent. In early February, $285B in SaaS market value evaporated in a single day as investors realized agentic AI doesn't need software seats. If AI agents replace the humans who held those seats, per-seat revenue models collapse. That's forcing every SaaS company to rethink what they charge for, and outcome-based pricing is one of the answers people keep reaching for. At the same time, the economics are rough. Bessemer warns that companies in "soft ROI territory," copilots that offer advice without closing the loop, face significant churn. 84% of enterprises report margin erosion from AI workloads. 80% miss their AI infrastructure forecasts by more than 25%. Outcome-based pricing is one way to prove you believe in your own product. But it also means absorbing the risk when your product doesn't deliver. There's a catch, though. The companies most pressured to adopt outcome pricing, the ones struggling to prove ROI, are usually the least equipped to absorb the cost risk. The companies that could absorb the risk don't need outcome pricing to prove their value. The renewal cliff doesn't push companies toward outcome pricing specifically. It pushes them toward proving value at all. Most teams haven't solved that yet. --- ## What's actually going to happen Outcome pricing stays narrow. It'll expand in categories where AI is autonomous and attribution is clean, but it doesn't become the default. The bigger shift is hybrid. Base subscription for predictability, outcome or usage components for value alignment. That's what Salesforce landed on. It's where Intercom, Cursor, and most companies end up. Growth Unhinged's 2025 survey found 41% of companies now use hybrid pricing, up from 27% a year ago. The pattern: | Company | Model | |---------|-------| | Intercom Fin | Base subscription + $0.99/resolution | | Salesforce Agentforce | Per-user license + Flex Credits ($0.10/action) | | Cursor | Monthly subscription + compute credits | And then there's the infrastructure question. Most teams can't ship outcome pricing even if they want to, because their billing stack doesn't support it. Stripe found that 92% of AI startups changed pricing post-launch due to inadequate billing systems. Metering outcomes requires different instrumentation than metering tokens. --- ## Ways to get outcome alignment without the risk This tends to get framed as binary: you charge for outcomes or you don't. But there are ways to get the alignment benefits without absorbing all the cost risk or arguing over attribution. **Prepaid outcome credits.** The customer buys a block of credits upfront, each redeemable for one outcome (a resolved ticket, a processed document, a completed case). The meter is outcome-aligned, but the cash mechanism is prepaid. The buyer knows their max spend. You know your cash flow. That's basically what Salesforce ended up doing with Flex Credits. **Graduated outcome pricing.** Instead of a flat fee per outcome, price in tiers based on complexity or compute required. A simple resolution costs one credit. A complex one that requires escalation or multiple model calls costs three. This lets you absorb some cost variance without eating the full spread between your cheapest and most expensive outcomes. **Base subscription with outcome-based expansion.** Charge a flat monthly fee that covers a set number of outcomes. Once the customer exceeds that threshold, additional outcomes are billed individually. The base gives predictability. The overage captures value at scale and naturally segments light users from heavy ones. **Performance bonuses on top of usage pricing.** Charge for usage (tokens, API calls, compute) as your baseline, then add a bonus when outcomes exceed a threshold. The customer pays a predictable usage bill, and you earn more when the product delivers measurably well. Instead of absorbing downside, you capture upside. None of these require solving attribution completely. They get you close enough while keeping billing simple enough to actually ship. --- ## If you're deciding right now Three questions before committing to outcome pricing: **Can your product complete the task autonomously?** If a human is in the loop, you're a copilot. Price on activity or access, not outcomes. **Can you define success in a way both you and the customer agree on?** If "resolved" means different things to you and your buyer, you'll spend more time arguing about invoices than improving the product. **Can you absorb the cost variance across your customer base?** One difficult case that burns 10x the compute of average will happen. If your margins can't handle it, outcome pricing will hurt you. If yes to all three - outcome pricing is real for you. If no to any of them, hybrid is the better path. Use outcome framing in positioning, but price on usage or workflows until attribution catches up. --- The goal was never outcome-based pricing specifically. It was aligning price to value. Outcome pricing is one way to get there, but for most companies right now, it's not the most practical one. **Key Takeaways:** - Outcome-based pricing works in narrow conditions: autonomous AI, clear completion signal, bounded cost variance - Most companies can't ship it - attribution is unsolved, operations are messy, billing stacks don't support it - The real shift is hybrid: base subscription + outcome or usage components. That's where most teams end up once the theory meets the invoice.