Most software is priced by the seat: a monthly fee for every person who can log in. In AI products, more of the work is done by the software itself, so the seat stops measuring what the customer gets.
On 10 November, Zendesk launched plans with no per-seat charge and an Outcome Budget for eligible outcomes. The customer funds a balance, and Zendesk deducts from it each time qualifying work is finished. Anyone on the team can be added at no extra cost. The launch post says conversational outcomes start at $0.99 (as of 14 Nov 2026); the full rate card is available only through Zendesk's sales team, and the plans are in an early access programme. This post explains why that model grows usage, how to test it from your customer's side, and how a small SaaS or AI product can copy it this month.
The idea and why it works
The play has two halves. First, drop the seat fee, so inviting a colleague costs nothing. Second, charge for a finished, checked piece of work (an "outcome") and let the customer set a budget that caps what they spend.
Removing seats grows usage because the person who decides whether to invite a colleague is usually the person watching the bill. When each invite costs money, teams share logins and keep the tool to one department. When invites are free, more people use it, more workflows depend on it, and it becomes harder to cancel.
Billing per outcome aligns price with value. The customer pays when the product does the job, and nothing for idle logins. Revenue then grows as your product gets better at the work.
Zendesk has evidence it is willing to bet on. Its chief executive, Tom Eggemeier, told TNW the company will end the year with about $400m of AI revenue, up from about zero two years ago, most of it from AI agents that customers pay for only when they resolve a case. He also said about 100 customers had tested pricing without seats. These are the company's own, unaudited figures.
Who it's for
This fits a product where the software does a countable job: resolving a support ticket, filing an invoice, publishing a post, booking a meeting, cleaning a dataset. It fits best when you can tell from your own data whether the job was done.
It fits poorly when value is hard to count (a design tool, a notes app) or when your costs scale with users rather than jobs. There, keep seats or use a usage ladder.
What you need
Tools: an event log that records each completed job with an ID, your billing system's metered or usage billing feature, and a simple dashboard customers can see.
Time: about a week to write the outcome definition and wire up metering, then two billing cycles of testing.
Cost: mostly your time. Set a minimum spend or platform fee to cover fixed costs.
Step by step
Pick one outcome. Choose the job customers judge you on, written as something finished and checkable, such as "an invoice matched to a payment and marked paid".
Write and publish the outcome definition. Say what counts, what never counts, and how you check it. Zendesk publishes its outcome definitions, including exclusions such as spam, test data and the source ticket in a merge. A public definition stops "resolved" becoming an argument.
Add a quality check before you charge. Count an outcome only after it passes a check: the customer confirms, no follow-up arrives within a set window, or an automated review passes. Zendesk says AI-handled resolutions are evaluated by an LLM for quality and completeness before they count.
Remove the seat limit. Allow unlimited users, with roles so admins stay in control.
Let the customer set the budget. At checkout, ask for a monthly outcome budget. Send alerts at 50%, 80% and 100%, and let the customer choose what happens at the cap: pause, keep working and notify, or top up. (Our post on bill-shock alarms covers the alert design.)
Show every charge. Give each billed outcome a line the customer can open, with the job, the time and the check it passed. Zendesk's reporting lets customers trace charges to individual outcomes.
Run it beside your current plans. Offer it to a small group first, keep seat plans available, and compare the metrics below.
Test it from the user's side
Before launch, read your pricing page as a customer and ask four questions. Am I paying for anyone who does not use it? (Should be no.) Can my bill go above the number I chose? (Should be no, unless I opt in.) Can I see exactly what I was charged for and why? Can I dispute a charge and get it back? Fix any wrong answer before you change prices. Intercom's Fin, a support agent sold on $0.99 per outcome as of 14 Nov 2026, gives a useful benchmark: it charges for at most one outcome per conversation, even if Fin takes several actions.
Template: pricing page copy and outcome definition
Pricing page block:
Unlimited teammates. Pay only for finished work.
Invite your whole team free. You pay [price] for each [outcome], and nothing for logins. You set a monthly budget; we will alert you at 80% and never charge above it without your approval. Every charge links to the job it paid for. Platform fee: [amount], which includes [n] outcomes.
Outcome definition (example for an invoicing tool):
An outcome is one invoice that our system matched to a payment and marked as paid, and that was not reversed within 7 days. Each invoice counts once, however many people touched it. We do not charge for: duplicates, test or sample invoices, invoices you imported already paid, failed or partial runs, or invoices you mark as wrong within 7 days (we refund these automatically).
Illustrative model (made-up numbers, not results): a five-person team pays $20 a seat, so $100 a month, and two people never log in. Under an outcome plan at $0.50 per paid invoice with a $20 platform fee, the same team processing 200 invoices pays $120, and all five people get access. If the product spreads to a second team and volume doubles to 400 invoices, the bill rises to $220 without a single seat sale. If volume drops to 50, the bill falls to $45, which is why the platform fee matters.
Metrics to watch
Active users per account: should rise once invites are free.
Outcomes per account per month: your new revenue driver.
Dispute or refund rate: the share of outcomes customers challenge. Rising disputes mean your definition is too loose.
Budget hit rate: how often customers reach their cap. Top-ups signal expansion; pauses signal a budget set too low.
Gross margin per outcome: your cost to deliver each outcome, including AI model costs.
Pitfalls
A vague definition. If "resolved" can mean the customer gave up, you will be charging for failures, and customers will notice.
Unpredictable bills. Finance teams want a fixed number; the customer-set budget gives them one.
Perverse incentives. Charging per ticket closed rewards closing fast. Add the quality check and a reopen window.
Revenue holes. Without a minimum spend or platform fee, a slow month can fall below your costs. Zendesk's lowest plan uses a minimum spend for this reason.
Model costs. If each outcome costs you more in AI calls than you charge, scale makes it worse. Check your cost per solved task first.
Variations
Hybrid: keep a small seat fee for power users and charge outcomes for the AI work. Our post on putting AI in every seat covers the seat-plus-top-ups version.
Tiered outcomes: price simple and complex jobs differently, building on the split Zendesk makes between conversational and operational outcomes.
Prepaid packs: sell outcomes in discounted blocks for a fixed invoice.

