.webp?width=1921&height=622&name=usecase-banner%20(1).webp)
Key takeaway
- Per-seat pricing punishes AI vendors for succeeding. How XenonStack is moving to platform, governance and outcome pricing with an automation dividend.
How should agentic AI be priced? By the work done and the outcomes delivered, not by seats or people. Per-seat and per-person pricing punishes AI vendors for succeeding: when an agent does the work, the customer needs fewer seats, so a vendor whose product works loses revenue at renewal. XenonStack is moving to platform, governance and outcome pricing, with a commitment that as its agents do more of the work, customers pay less every year.
The short version
- Per-seat and per-person pricing punishes AI vendors for succeeding.
- XenonStack is moving to platform, governance and outcome pricing.
- The commitment: as agents do more of the work, customers pay less every year.
The Question That Changed the Proposal
A few weeks ago, the XenonStack team was finalising a managed services proposal for a bank. The pricing sheet looked like every proposal in the industry: roles, rates, headcount, annual escalations.
Then came a simple question. If our agents take over half of this work by Year 3, who benefits?
Under that contract, the honest answer was: nobody. The bank would keep paying for ten people. XenonStack would keep billing for ten people. And the AI built to do the work better would sit inside the proposal as a cost, not a promise.
That was the moment to change how XenonStack prices its work.
If our agents work, our customers should pay less. And we should still make more.
What Is the Per-Seat Trap?
Most enterprise software is priced per seat. Most services are priced per person. Both made sense when humans did the work and software helped them do it.
Agents break that logic. When an agent resolves the ticket, reconciles the account or triages the alert, the customer needs fewer seats and fewer people.
A vendor whose product works perfectly loses revenue at renewal. A vendor whose product works poorly keeps it.
Customers feel that perverse incentive. It shows up as AI features that never quite reduce headcount, pilots that never quite reach production, and contracts that grow every year while the work gets easier.
Julien Bek at Sequoia made the point well in "Services: The New Software": in every profession, companies spend far more on getting the work done than on the tools that help. By his estimate, services spending runs at about six times software spending. The vendors who win the next decade will be paid for the work, not the seat.
What XenonStack Is Changing: Three Pricing Layers
The new pricing model has three layers, each tied to a budget customers already have.
| Layer | Priced on | What it covers |
|---|---|---|
| 1. Platform | Work done | The decision infrastructure, ElixirData, and the agent platform, ElixirClaw, are priced on decisions governed and agent work completed, not logins. Revenue grows when agents do more, not when more people sign in. |
| 2. Governance | Per agent | Every agent that touches regulated data must be governed, audited and reversible. MetaSecure and AgentGRC are priced per agent or per protected system, from the risk and compliance budget. |
| 3. Outcomes | Per result | Where a customer already outsources work, XenonStack takes on the outcome itself: a platform that runs, incidents resolved, evidence delivered. Priced per outcome, with service levels it stands behind. |
The outcome layer carries one more commitment: the automation dividend. Every year, as agents handle more of the work, a fixed share of that gain comes back to the customer as a lower price. XenonStack keeps the rest as margin.
Our margin should come from automation, not from rate cards.
What Does This Look Like in Practice?
Back to that bank. The traditional proposal priced a team of engineers over a multi-year term to support an open-source data platform. The new structure prices what the bank actually needs.
| Before | After |
|---|---|
| Pay for 10 people every month | Pay a fixed run fee for a platform that stays up |
| Rates rise every year | The price falls every year through the automation dividend |
| Tickets are effort | Each resolved incident is a priced outcome, with service levels |
| Penalties are rare and vague | A clear bonus or refund against availability targets |
| Knowledge leaves with the people | Every resolution is recorded and owned by the bank |
Behind it, agents handle alert triage, runbooks and routine fixes. Engineers take the exceptions and the hard problems. Every decision, by an agent or a person, is captured with its context, so the next one is faster.
One number is measured above all: the share of work that still needs a human. It should fall every quarter. When it does, the bank pays less and XenonStack earns more. The vendor's incentives finally point the same way as the customer's.
Four Questions for Your Next AI Contract
If you are a CIO, CFO or COO renewing AI and services contracts this year, ask every vendor these four questions.
- Does your price fall as your automation improves? If not, you are funding their margin with your efficiency.
- What exactly am I paying for? Logins and hours are inputs. Insist on outcomes you can measure.
- Who owns the decisions your agents make in my business? That record is your institutional knowledge. It should stay with you if you switch vendors.
- Can you explain any single decision? In banking, insurance, healthcare and government, an outcome you cannot audit is a liability, not a saving.
The Test
This shift is not easy for a company like XenonStack. Rate cards are predictable. Outcome contracts move risk onto the vendor, and they expose whether the technology really works.
That is exactly why it is worth doing. If we believe our agents can do the work, we should be willing to be paid for the work.
Seats measure attention. Outcomes measure value.
See How Outcome Pricing Would Apply to Your Work
In a Context Workshop, XenonStack works with your operations and finance leaders to pick one outsourced or run workload, define the outcomes and service levels that matter, and map which decisions agents can take on. Related reading: AI Adoption vs Absorption.
Book a Context Workshop | Explore ElixirClaw Agentic OS
Frequently Asked Questions
Why does per-seat pricing fail for AI agents?
When agents do the work, customers need fewer seats and fewer people. A vendor priced per seat loses revenue when its product works well, which rewards products that never quite reduce the workload.
What is outcome-based pricing for agentic AI?
The customer pays per result, such as a platform that stays up or an incident resolved, with service levels attached, instead of paying for logins or hours.
What is the automation dividend?
A commitment that each year, as agents handle more of the work, a fixed share of that gain is returned to the customer as a lower price while the vendor keeps the rest as margin.
What are XenonStack's three pricing layers?
Platform priced on work done, governance priced per agent or protected system, and outcomes priced per result.
What should buyers ask AI vendors at renewal?
Whether the price falls as automation improves, what exactly is being paid for, who owns the decisions agents make, and whether any single decision can be explained.
Source
Julien Bek, Sequoia, "Services: The New Software".
This article first appeared in Navdeep Singh Gill's LinkedIn newsletter, AI + Human = Human Squared.
Explore the XenonStack Platform and Agent Network
Platforms: ElixirData · ElixirClaw · ElixirHub · NexaStack · MetaSecure · Neural AI · AutonomousOps · CohortIQ · AI Aviator · SAIF Aviator
AI agents: AgentAnalyst · AgentEvaluation · AgentForce · AgentGRC · AgentHR · AgentInstruct · AgentLabel · AgentPixel · AgentQA · AgentRAI · AgentScrum · AgentSearch · AgentSketch · AgentSRE
XenonStack: xenonstack.com · xenonstack.ai · xenonstack.ae · xenonstack.jobs
