Investors
The layer enterprise AI is missing.
Archiveris builds the runtime that holds a company’s business definitions — its data model, its decisions and its permitted actions — as governed, versioned infrastructure that applications and AI agents call instead of rebuilding.
The thesis
Business logic was never made infrastructure.
Compute, storage, payments and identity were all extracted from applications and turned into services that everything calls. The business itself never was. What a claim is, when a payment may be released, who can authorise an exception — that knowledge still lives inside each application, each integration, and now inside each prompt.
Generative AI made producing software cheap. It did not make a business definition trustworthy. A model can give a plausible account of an expense policy. It cannot be relied on to give the same account twice, or to show the clause it came from.
So the constraint has moved. The question is no longer whether software can be built quickly. It is whether what the software does can be governed.
What the runtime holds
Three governed building blocks, composed into operations that applications, workflow tools and agents invoke as one call.
- Jiwari — what is true: entities, relationships and state
- Kebiki — what follows: policy compiled into deterministic decisions
- Teban — what may happen: permitted actions, authority and evidence
- Business Operations — the composition the three are called through
Orchestration stays with the customer’s own workflow engine, BPM suite or agent framework. We compose; they orchestrate.
Why now
Four things changed at once.
This company would not have been buildable, or saleable, three years ago. The conditions that make it both arrived together.
Pilots are moving into production, and what stops them is rarely model capability. It is accountability: what the agent may do, under whose authority, and with what evidence that it was entitled to.
When any team can produce a working application in a week, the application stops being the scarce asset. The definition it is built on becomes the thing worth owning.
Tool calling and MCP standardised how systems expose capability to AI. A governed runtime now has a distribution surface into every agent framework that did not exist two years ago.
Boards, auditors and regulators are putting to automated decisions the question they have always put to people: on what basis? A system that cannot answer gets withdrawn, however well it performs.
What we have built
Three products that sell on their own.
Each solves a problem by itself, which is how we get in. Composition is why a customer who starts with one ends up with the runtime.
Jiwari
A governed model of the business world — entities, relationships, state and context — that applications and agents read instead of each inventing their own.
Kebiki
Policies, contracts and regulations compiled into tested, versioned decision APIs that return the same result every time, with the rule and the source passage behind it.
Teban
What an actor may do in the current situation, with the conditions, limits, authority and evidence each action requires — the boundary an agent acts inside.
Business Operations
The composition layer: a published, deterministic function assembled from the blocks a job needs, invoked over REST, MCP and SDKs under one contract and one trace.
Kebiki is the commonest front door: a decision is legible to a buyer in a way an architecture is not.
Why it compounds
The second operation is cheaper than the first.
Every definition a customer adds raises the value of the ones already there. A customer model built for claims serves servicing and underwriting. An eligibility decision called by an agent today is called by an application tomorrow. The first operation has to justify itself; the tenth is mostly assembled from what already exists.
What accumulates is not a stack of configuration. It is the authoritative account of how that business works, versioned and depended on by the systems around it. Enterprises replace that reluctantly, and they do not replace it in a quarter.
Where the category comes from
We are writing the discipline, not only the product.
The reference architecture Kebiki and Teban implement is published as a book, Policy Engineering, written by our founder and stewarded by a non-profit Institute as an open body of knowledge. It is free to read.
Defining the vocabulary a category uses is a slower advantage than a feature, and a more durable one. It is also how we reach the architects who decide what gets built on.
The model
How the company makes money.
Two motions, pointed at the same account. Services prove the first use case; the platform is what the customer keeps paying for.
| Motion | What it is |
|---|---|
| Platform | Subscription to the runtime: the products a customer uses and the operations they publish on them. |
| Professional services | Engagements that take one use case to production and leave governed assets the customer’s own team runs. They pull platform adoption and keep us inside the problems. |
| Who buys | Architecture, risk and the people accountable for what automation does — the ones who have to answer for a decision after it is made. |
Everything else is a conversation.
This is a public page, so it carries no financials, pipeline or roadmap. Those we share directly, under NDA. If you invest in enterprise infrastructure and the thesis above is one you recognise, write to us.
Nothing on this page is an offer to sell, or a solicitation of an offer to buy, any security. It describes the company’s strategy and architecture, not its performance.