For as long as you've bought software, you've paid for access. Seats, tiers, monthly plans — the bill arrived whether the tool did anything for you or not. That's starting to change.
A growing set of AI tools now bill on outcomes: you pay only when the agent finishes the job. Intercom's Fin support agent set the pattern at 99 cents per resolution. Zendesk followed with roughly $2 per automated resolution pay-as-you-go, or about $1.50 if you commit to volume. And this spring HubSpot moved two of its Breeze agents over: the customer agent dropped from $1 per conversation to 50 cents per resolved conversation, and the prospecting agent went from a recurring monthly fee to $1 per lead it recommends for outreach. Both come with a 28-day free trial.
Read that pricing again, because the shape of it matters more than the numbers. A conversation the AI touches but can't finish: free. A lead it researches but doesn't surface: free. You're not renting a tool anymore. You're paying a contractor an invoice for completed work.
Why this is a big deal for a small business
Seat-based pricing always favored big companies. A $500-a-month platform is a rounding error at 200 employees and a real decision at four. Outcome pricing flips that: if the agent resolves 50 conversations for you this month, you pay for 50. If it resolves 5,000 for someone else, they pay for 5,000. The small operator finally pays small-operator prices for the same software.
It also makes the buying decision honest in a way software rarely is. You don't have to guess whether you'll "get value" out of a subscription. You can do contractor math: what does it cost you today — in your time, or a VA's, or an employee's — to handle one support conversation or chase one lead? If your answer is more than 50 cents, and it almost certainly is, the agent earns its line on the bill. If the agent can't finish the work, the bill is zero and you've learned something for free.
A seat is a bet that you'll use the software. An outcome is a bet that the work gets done — and you only pay when it does.
The fine print is the whole game
Before you get excited, know that every vendor defines the billable word differently — and the definition is where your money goes.
"Resolved" is the vendor's word, not yours. Intercom counts a resolution, a procedure handoff, or a disqualified conversation as a billable outcome. Zendesk's definition of an automated resolution is different again. A bot that closes a chat isn't the same as a customer whose problem is actually solved — so watch your reopen rate and what customers say afterward, not just the resolution count.
Some outcomes cost more than others. On Intercom, a standard outcome is 99 cents, but a lead-qualification outcome is $9.99. On HubSpot, a recommended lead is $1 whether or not that lead ever replies. "Recommended" is not "closed."
Minimums and platform fees still exist. Intercom has a floor of around 50 resolutions a month, and HubSpot's agents require a Pro or Enterprise plan underneath them. The outcome price is a line on the bill, not the whole bill.
The twenty-minute check
Here's the math to run before any trial. Count last month's support conversations — your inbox, chat widget, and DMs all count. Estimate how many were routine: order status, hours, how-do-I questions, password resets. Multiply that routine number by the per-resolution price. Then compare it to what those same conversations cost you now, at whatever you'd honestly bill for the hours.
If the gap is wide — and for most service businesses it is — start a trial. But write down the vendor's definition of "resolved" first, and check your reopen rate at day 28 before the billing starts. The trial isn't for testing whether the AI answers. It's for testing whether the vendor's definition of finished matches yours.
Software billed like software for forty years. Agents are starting to bill like labor. When the bill scales with work actually completed, a four-person business gets access to capacity it never could have hired — and a very clear way to check whether it's worth it.