The usage-based vs outcome-based pricing debate almost always gets framed as a question of what your customers want. That's the wrong frame. Customers nearly always say they want to pay for outcomes, results, not effort, and if you ask them, they'll tell you outcome pricing every time. The binding constraint isn't preference. It's what you can measure and bill reliably, and most AI companies can't yet prove or cost an outcome. So the real question isn't which model sounds fairest; it's which model your product can actually support today, and what bridges you to the next one. Here's a decision tree built on the two axes that actually decide it.
The three models in brief
Each model is covered in depth in the complete playbook on pricing AI agents; here's the one-paragraph version of each so the comparison has footing.
Usage-based pricing charges per token, call, or agent run. It's the easiest model to start with because cost and revenue move together, a heavy user automatically pays more. Its risk is mundane: under-pricing the unit. Roughly 61% of software companies now run some usage component, which tells you it has become the safe default rather than the bold choice.
Outcome-based pricing charges per result, a resolved ticket, a qualified lead, a dollar saved. It's the most aligned to value and removes the buyer's risk entirely, which is why Intercom ($0.99 per resolution), HubSpot ($0.50 per resolved conversation), and Zendesk (up to $2.00 per automated resolution) have moved toward it. Its risk is structural: a variable, uncapped cost per outcome that you can only manage if you can prove and cost the result. Only about 9% of companies have fully implemented it.
Hybrid pricing combines a base fee with a usage or outcome layer. It's now the dominant real-world model, somewhere between 41% and 61% of companies depending on the survey, because it gives procurement a fixed number to sign while capturing upside as customers scale. Its risk is the quietest: the base fee can hide a variable layer that's losing money.
The two axes that actually decide it
Forget customer preference for a moment. Two questions determine which models are even available to you.
The first axis: can you prove the outcome? Is there a clean, mutually-agreed signal that the result actually happened, a ticket genuinely resolved, a lead genuinely qualified, a dollar genuinely saved? If you and your customer can't agree on a definition of "resolved" you both trust, outcome pricing degrades into a standing dispute, no matter how much the buyer says they want it.
The second axis: can you cost the outcome? Do you know your fully-loaded cost per outcome, including the attempts that failed and billed nothing? If you don't, you're quoting a fixed price against an uncapped cost, and you won't find out until the renewal. (This is the trap covered in detail in is your outcome-based pricing actually profitable.)
Outcome pricing is only safe when both answers are yes. If you can meter usage cleanly but can't yet prove or cost an outcome, usage or hybrid is your honest answer. If you can't even meter the unit reliably per customer, you have a measurement problem to solve before you have a pricing decision to make.
The decision tree
Run your product down this path and stop at the first honest "no." 1. Is your agent autonomous, does it act without a human bounding its usage? If no, it's a copilot a person drives, and a seat or seat-plus-usage hybrid can still work; predictability is worth more than perfect value capture here. If yes, continue. 2. Can you cleanly meter the unit of usage, tokens, calls, or runs, per customer? If no, fix metering first; run a simple usage or credit model in the interim. If yes, continue. 3. Can you prove the outcome with a success signal both sides accept? If no, use usage-based or a hybrid of base plus usage. If yes, continue. 4. Can you cost the outcome fully, including failures, and does your price beat your cost across customer segments, not just on average? If no, use a hybrid with a usage floor to protect margin while you build cost instrumentation; do not go pure outcome yet. If yes, you've earned outcome-based pricing, or a hybrid with an outcome layer on top.
Most companies discover they belong one or two steps earlier than they hoped. That's not a failure, it's the difference between a pricing model you can defend and one that defends itself right up until it bankrupts you.
The bridge: credits and hybrids
The path from usage to outcome runs through hybrids, and the on-ramp is usually credits. Credit-based packaging has surged, up about 126% year over year, because it solves both sides of the transition at once: the buyer pre-pays a predictable block, and you book committed revenue while you learn real usage patterns. Hybrid pricing is the dominant transition state in 2026 for the same reason. A fixed base satisfies procurement's need for a purchase-order-friendly number, and the variable layer lets revenue grow with value without a renegotiation.
Treat the move from usage to outcome as a graduation, not a launch. You earn each step by proving you can measure and cost what you're about to charge for.
Usage-based vs outcome-based pricing: the profitability tiebreak
When two models both seem to fit, the tiebreaker isn't which one customers prefer. It's which one's margin you can see and defend as you scale. Usage margin is the easiest to protect, because cost and revenue rise together. Outcome margin is the hardest, because the cost is uncapped. Hybrid can quietly hide either problem behind the base fee.
Here's the part most comparisons miss: the instrumentation you need to answer "is this profitable?" is identical across all three models. Cost and margin per unit, per outcome, and per customer is the same measurement problem whether you bill for tokens, results, or a blend of both. Which is exactly why you build that visibility first and choose the model second, not the other way around.
Pick the model your product can prove and cost today. Bridge to the next one with credits or a hybrid. And never sell a model whose margin you can't watch in real time. The companies that get this right aren't the ones with the cleverest pricing page, they're the ones who can tell you, per customer, whether each model is making money.
Paygent lets you design and launch usage, outcome, or hybrid pricing, and shows the true cost and margin under each, so you can choose with numbers instead of vibes, and switch models without flying blind.
