Pricing
Pricing
Three tiers built around the AI systems you connect and the activity they generate: platform, team and enterprise.
No numbers on this page
Not a table and not a “from” figure — that is the version of a made-up price which looks most like disclosure while committing to nothing.
What is here instead is exactly what you are metered on, and the six things you are not.
Never metered
6
Prices listed
0
The useful half
Six things you are never charged for, and why each would be the wrong meter.
What a governance product meters on is a design decision with consequences, and most of the obvious choices are actively harmful. This list is checkable, and it is the part you need in order to model your own bill.
- Decisions evaluatednever metered
- Metering evaluations would make the cheapest configuration an ungoverned one. For a product whose value is that every action passes the boundary, charging per action is an incentive pointed at exactly the wrong outcome.
- Refusals and held actionsnever metered
- You would be paying more for the control working. A refusal is the product doing its job, and it costs us the same as an authorization.
- Seatsnever metered
- The things being governed are not people. How many humans open the console has no relationship to the value or to our cost, and per-seat pricing would discourage exactly the people who should be looking — approvers and auditors.
- Policies writtennever metered
- A per-policy charge would discourage writing the narrow, specific rules that make governance work, in favour of a few broad ones that do not.
- Evidence retainednever metered
- Retention is usually an obligation rather than a choice. Charging for it would put a price on meeting one, and the storage is not the expensive part.
- Model callsnever metered
- OpsAI is not on the inference path and does not proxy your models. There is nothing here to mark up, and a product that charged for it would have an incentive to insert itself.
A meter that punishes evaluation would make the cheapest configuration an ungoverned one.
That is the whole reason this list exists. In the sample estate, 240 decisions include 20 held for a person and 5 refused — and charging for any of those would mean charging more the better the control worked.
IllustrativeIllustrative activity from the OpsAI sample estate. The decision record.
What it is metered on
The AI systems you connect, and the activity they generate.
Two dimensions, both of which grow with the value rather than against it. Connecting a system under governance is the thing worth doing, and it is also the thing that moves the number — which keeps the incentive pointed the same way as yours.
- AI systems connected
- Registered and under governance. The sample estate has 11. A system found by discovery and not yet onboarded is not counted — you are not charged for finding a problem.
- Activity they generate
- In bands rather than per action, so an unexpected busy week is not an unexpected invoice. Bands are the mechanism that keeps this from behaving like a per-decision meter.
- Discovery is not metered
- You are not charged for the systems discovery finds, only for the ones you decide to govern. Charging for findings would make the honest thing expensive.
Why bands rather than per action
Because per-action billing has the same defect as per-decision metering, arriving more slowly. A team that watches a counter increment learns to route around the boundary for the high-volume paths, and the high-volume paths are usually the ones worth governing.
A band has a ceiling you can plan against and no per-request signal to react to. It is a worse revenue model and a better control.
Three tiers
Described by who they are for, not by a feature matrix.
And with one commitment that matters more than the shape: no governance capability is held back for a higher tier. A control you cannot afford is not a control, and a policy engine that refuses to enforce until you upgrade would be an unusually cynical product.
- 01
Platform
quoted- Who it is for
- One team bringing the AI it already runs under governance.
- What it changes
- Scoped to the AI systems that team connects. Everything in the product is present — there is no governance feature held back for a higher tier, because a control you cannot afford is not a control.
- 02
Team
quoted- Who it is for
- Several teams, each owning its own policies.
- What it changes
- Distributed policy ownership is the point rather than an add-on. Scales with the AI systems connected across those teams, not with the number of people involved.
- 03
Enterprise
quoted- Who it is for
- An estate, with a deployment requirement and an assessor.
- What it changes
- Adds the deployment options where nothing leaves your network, and the operational commitments that come with running it yourself or in your own cloud account.
Tiers
3
platform, team, enterprise
Metered dimensions
2
systems and activity
Controls behind a paywall
0
every tier enforces
Prices on this page
0
see below
The obvious question
So why is there no number?
Because there is not one to publish yet, and the alternatives are all worse than saying so. This is the same position as the certification status and the empty changelog — the honest answer is available and a plausible one would be an invention.
- A “from” price would be worse than none
- It looks like disclosure and commits to nothing. Everybody who has bought enterprise software knows the figure has no relationship to the eventual invoice.
- A full table would be an invention
- Numbers nobody has committed to, on a page a reader would reasonably plan a budget against. That is a more consequential fabrication than most on a website.
- What you can get today
- A number for your estate, from a conversation about how many systems you would bring under governance and which deployment option you need. The two metered dimensions above are the whole input.
Where to start
Count the AI systems you would put under governance first.
That number, plus roughly what they do and which deployment option you need, is the entire input. It is also worth counting before any pricing conversation, because most estates find the answer is larger than expected.