Skip to main content

What is The One

The One is the AI Governed Layer: the layer that sits between your source systems and your AI. It resolves every source into one governed, canonical record — a golden record — per real-world entity, and serves that record to applications over a REST API and to agents over an MCP endpoint, with the caller's own permissions enforced on every request, at low latency, and fully audited.

People reach the same record through a web console, under the same enforcement.

That is a category, not a feature list — four principles a platform has to satisfy. What the AI Governed Layer means sets them out.

The problem

The console home screen The console — one of the three ways a golden record is served, alongside the API and the governed endpoint.

Every organization of any size has the same one. The customer in the CRM, the account holder in the billing system, and the caller in the support tool are the same person, and no system knows it.

The consequences compound:

  • Reports count that person three times.
  • A privacy request has to be executed three times, and one gets missed.
  • A change of address updates one system.
  • An AI assistant asked "how many customers do we have?" answers confidently and wrongly.

No amount of downstream cleverness fixes this. The resolution has to happen once, in one governed place.

What the platform does

CapabilityWhat it means
ModelDefine the entity types and attributes you master
StandardizeTransform incoming values into canonical form
ValidateConstrain what is acceptable, and score what arrives
MatchIdentify records describing the same real-world thing
MergeConsolidate them, reversibly, with lineage preserved
RelateModel governed links between entities
GovernEnforce access control and masking on every read
ServeExpose all of it to applications, agents, and people

What makes it different

Matching weighs evidence. Rather than evaluating fixed rules, the engine scores how much each agreement and disagreement actually tells you, including how surprising a specific value is. Agreeing on a rare surname is strong evidence; agreeing on a common one is nearly none, and the score reflects that without configuration.

Governance is architectural, not bolted on. Access control is enforced in four independent layers, the innermost by the database. Masking is applied when a response is generated, so no endpoint can return what a caller may not see.

AI is a first-class consumer, not an afterthought. Agents reach the platform as the signed-in user and are subject to identical enforcement — the principle the AI Governed Layer is built on.

Nothing is destroyed. Merges reverse. History is complete. The audit log is append-only and its gaps are detectable.

Who it is for

RoleUses it to
Data architectModel the domain and configure matching
Data stewardResolve ambiguous matches and fix data
Compliance officerManage consent and answer privacy requests
Integration engineerLoad data and consume golden records
AdministratorManage users, roles, and access policy

Next


Last verified against commit 952ed91b (2026-08-03)