Per-seat pricing made sense when software helped people do work. It makes less sense when software does the work. If an AI agent handles the invoices, charging per user means the vendor earns less as the agent gets better, and the buyer pays for access rather than results. Outcome pricing fixes that, but only where the outcome is clear-cut. For everything else, a floor fee plus a capped share of verified gains is the version that holds up.
Why seats stop making sense
A seat is a proxy for value. The assumption is that each user gets a roughly fixed amount of benefit from the software. That breaks down with agents in two ways.
First, the value moves away from the user. If an agent processes 2,000 invoices a month with one person reviewing exceptions, the value is in the 2,000 invoices, not the one seat.
Second, the incentives flip. A vendor paid per seat does better when the customer needs more people. A vendor paid per outcome does better when the agent resolves more work. Only one of those lines up with what the buyer wants.
The big platforms are already splitting the difference. Microsoft still sells Microsoft 365 Copilot per user, but prices Copilot Studio agents by consumption, at $0.01 per credit or $200 a month for 25,000 credits. That is metered usage, not outcomes, but it shows seats alone no longer fit agent work.
Where outcome pricing works
Outcome pricing has grown fastest where the result is binary. A support ticket is resolved or it isn't. A payment is completed or it isn't. Attribution is simple when the product does the whole job.
Customer service shows how far this has gone. Published per-resolution prices now run from about $0.50 to $2.00, and HubSpot cut its Customer Agent price to $0.50 per resolved conversation in April 2026, down from $1.00. That is good for buyers of support automation, and a warning for anyone selling it: once an outcome is standard and easy to measure, its price falls fast.
The lesson for back-office work is to price outcomes that are worth real money and hard to fake: a policy checked line by line, an invoice matched and posted, a renewal file complete. Those are worth tens or hundreds of dollars each, not cents.
Where it doesn't
Pure outcome pricing struggles when results depend on many hands. Faster cash collection depends on the provider's follow-up, but also on the client's invoicing accuracy, their customers' finances and the economy. Paying the provider for every day of DSO improvement, with no floor, invites arguments about who caused what.
HFS Research's data shows how early the market is. Only 17% of service contracts materially use outcome-linked pricing and 19% use gain-share, yet about 80% of leaders expect outcome pricing to grow. HFS calls gain-share the bridge to get there.
The hybrid that holds up
The structure we use has three parts:
- A floor fee that covers the cost of running the service: platform, models, the exception desk and governance. It keeps the service funded in a slow month and stops the provider cutting corners.
- An outcome component. For standard work, a price per verified transaction. For work where value depends on context, a share of verified gains above an agreed baseline.
- Guard rails. Written outcome definitions, a calibration period, and an annual cap so the buyer's maximum spend is known.
The part that makes it trustworthy is the word "verified". Every billed item needs evidence behind it, or the model turns back into paying for activity with better branding.
See how outcome pricing works at Agentic MSP, or read how to write an outcome definition.