# AI Agent Pricing for Small Business: Paying per Seat vs per Outcome

- Canonical: https://spiai.ai/blog/ai-agent-pricing-per-outcome-small-business
- Published: 2026-11-01
- Author: SpiAI
- Description: AI agent pricing is moving from seats to outcomes. How per-resolution billing works, what counts as an outcome, and how a small business can compare.

---
On October 27, [Greg Isenberg posted](https://x.com/gregisenberg/status/2115172540238979474) a four-line prediction that passed 200,000 views and 500 replies within two days: "saas is becoming free / give away the crm / give away the dashboard / give away the app / and charge for the agent that does the work."

The replies split the way you would expect. Some compared it to printers, cheap hardware and expensive ink. Others pointed out that it is already happening in customer support, where vendors bill per resolved conversation instead of per agent seat. Skeptics said the platform still holds the value and the bill just moves to a different line.

For a 10 to 75 person service business, this is not an abstract debate. You are about to see more software priced by outcomes, and you may be asked by your own customers to price your work that way. Both sides of that change come down to one question: who gets to define what counts as done?

## TL;DR

- AI software pricing is shifting from paying per user seat to paying per outcome, such as a resolved support conversation.
- Intercom, for example, lists its Fin AI agent at $0.99 per outcome and charges at most once per conversation. Zendesk bills per verified automated resolution.
- Outcome pricing is only fair if the outcome is defined in a way you can audit. "The customer did not reply" is not the same as "the problem was solved."
- Set a monthly cap, sample the billed outcomes, and keep a person on the cases that matter. AI drafts, your team approves.

## What is per-outcome AI agent pricing?

Traditional software charges for access. You pay per seat, per month, whether anyone gets value or not. Per-outcome pricing charges when the AI completes a defined unit of work: a resolved support conversation, a qualified lead, a booked appointment, a processed invoice.

Customer support is the clearest live example. Intercom's help center describes Fin's pricing like this:

| Outcome type | Price | How Intercom defines it |
|---|---|---|
| Resolution | $0.99 | No further help is requested after the last AI answer |
| Procedure handoff | $0.99 | Fin runs a configured procedure that ends in a handoff |
| Disqualification | $0.99 | Fin decides a prospect does not match your criteria |
| Qualification | $9.99 | Fin matches a prospect to your criteria and routes them |

Intercom says it charges for at most one outcome per conversation, does not bill when Fin only answers a greeting, and deducts a resolution if the customer later comes back to the same conversation. It also lets customers set usage reminders and hard limits. Its pricing page notes that Intercom plans still combine seats for human teammates with usage for Fin.

Zendesk takes a different line. According to a recent comparison by Macha, Zendesk bills only "verified" automated resolutions, at about $1.50 each on committed volume or $2.00 pay-as-you-go, and does not bill conversations that were merely contained.

## Why does the definition of an outcome matter so much?

Look at Intercom's resolution rule again: the customer either confirms the answer helped, or leaves without asking for more help. A customer who got a perfect answer and a customer who gave up in frustration can look the same in that data. Macha's comparison makes exactly this point: the same 2,000 "resolutions" on two platforms are not the same 2,000 events.

That is not a criticism of one vendor. It is the central weakness of outcome pricing everywhere. The vendor measures the outcome, the vendor bills for it, and the vendor's definition decides how often you pay. When you evaluate an AI agent priced by outcomes, the price per outcome is the least important number on the page.

## How to compare seat pricing and outcome pricing

Use a simple worksheet before you sign anything.

1. **Count your real volume.** Pull three months of support conversations, inbound leads or invoices, whatever the agent would handle. Use your numbers, not the vendor's calculator defaults.
2. **Estimate what share the agent will actually finish.** Ask for the vendor's typical resolution rate for businesses like yours, then assume less until your own pilot proves otherwise.
3. **Write down the outcome definition word for word.** Ask what triggers a charge, what does not, and what happens if the customer comes back.
4. **Price both scenarios.** Seat plan cost per month versus outcomes times price, plus any platform minimum. Run it at your slow month and your busiest month.
5. **Add the cost of review.** Someone still reads escalations, checks a sample of resolved conversations and fixes the knowledge base. That time is a real cost under either model.
6. **Set a hard cap.** If the vendor supports usage limits, set one at a number you are comfortable paying before anyone reviews the results.
7. **Audit a sample every month.** Read 20 billed outcomes. If a meaningful share were not really resolved, you are paying for silence.

## When outcome pricing makes sense for a small business

Outcome pricing tends to work when the work is high volume, repetitive and easy to verify. Answering "what are your hours" or "where is my invoice" a few hundred times a month is a good fit. So is first-pass lead qualification where the criteria are written down.

It works badly when outcomes are rare, high stakes or hard to verify. A dental practice handling insurance questions, a law firm triaging new matters or a contractor quoting jobs should not let a vendor's definition of "resolved" decide when a customer is done. In those cases, a predictable seat price with the AI drafting and a person approving often costs less once mistakes are counted.

## Should your own business charge by outcome?

The same shift is reaching service businesses from the other side. As AI shortens the time a task takes, clients start asking why they pay by the hour. Some agencies, bookkeepers and consultants are moving to fixed or outcome-based fees.

If you consider it, borrow the lessons from software pricing:

- **Define the outcome in the contract.** "Monthly books closed and reconciled by the 10th" is auditable. "Better finances" is not.
- **Price the review, not just the production.** AI can make the draft cheaper. The judgment and accountability you sell are not cheaper.
- **Keep a cap and a floor.** Protect the client from surprise bills and yourself from scope creep.
- **Do not promise what AI decides.** If an outcome depends on an AI answer being right, a person on your team still signs off before it counts.

## The skeptics have a point

Isenberg's post is a prediction, and the replies were right to push back. Seat fees have not disappeared: Intercom still charges seats for human teammates alongside Fin usage. Platforms with your data in them still have pricing power. And outcome pricing can rise quickly if the agent gets good and the volume grows.

The practical takeaway for an owner is narrower than "SaaS is dying." It is that you will increasingly be quoted a price per unit of work, and you should treat that quote like any other contract with a variable cost: understand the unit, cap the exposure and check the bill.

## Verdict

Verdict: Hold until you can audit it. Per-outcome AI agent pricing is a good deal for high-volume, low-stakes, easy-to-verify work, and a risky one where "resolved" is hard to check. Before switching, run three months of your own volume through both pricing models, get the outcome definition in writing, set a hard monthly cap and read a sample of billed outcomes every month. Keep a person approving anything that commits your business to a customer. AI drafts, your team approves, and you pay for work you can see was done.

**Not sure whether an AI agent would pay for itself?** In a free 30-minute assessment we look at your real volumes, estimate what an agent would actually finish and show you where a person still needs to approve the work. [Book a free assessment](https://spiai.ai/)

## FAQ

### What is outcome-based pricing for AI agents?

It means you pay when the AI completes a defined unit of work, such as a resolved support conversation or a qualified lead, instead of paying a fixed fee per user seat each month.

### How much does an AI support agent cost per resolution?

Prices vary by vendor and change often. Intercom lists Fin at $0.99 per outcome, with lead qualification at $9.99. Zendesk bills per verified automated resolution, reported at about $1.50 committed or $2.00 pay-as-you-go. Check current pricing before you buy.

### Is per-outcome pricing cheaper than per-seat pricing?

It depends on your volume and how the vendor defines an outcome. Run your own three months of data through both models, including the staff time spent reviewing the AI's work.

### What counts as a resolution?

Each vendor defines it differently. Intercom counts a resolution when the customer confirms the answer helped or leaves without asking for more help. Zendesk bills only resolutions its verification step confirms. Read the definition word for word.

### How do I stop an AI agent bill from running away?

Set a hard usage limit if the vendor offers one, review usage weekly during the first month, and audit a sample of billed outcomes every month to confirm the work was really done.

### Should my service business switch to outcome-based fees?

Consider it for work with clear, auditable deliverables, such as monthly reconciliations. Define the outcome in the contract, keep a cap and a floor, and price the review and accountability, not just the time saved.

### Will SaaS really become free?

Not entirely. Some vendors are bundling software and charging for agent work, but seat fees still exist for human users and platforms with your data keep pricing power. Expect more mixed pricing, not free software.