The enterprise needs several sources of truth — and a way to relate them

The enterprise needs several sources of truth — and a way to relate them

Enterprise knowledge is distributed across architecture repositories, CMDBs, API catalogues, identity platforms, decision records and policy libraries. This is not always a defect.

Different systems can be authoritative for different aspects of the same environment. The architecture repository may describe intent. The CMDB reflects deployed operational reality. The API catalogue publishes interfaces. Identity systems govern access. Decision records preserve rationale. Policies define constraints.

The problem appears when a change crosses those boundaries and the organization cannot relate their information without losing meaning, ownership or provenance.

Every source was created for a reason

An architecture model, CMDB record, interface definition and decision paper do not describe the same thing badly. They describe different aspects of the environment for different operational purposes.

Some fragmentation is accidental. Some is architectural. Specialized sources exist because different teams need authoritative information for different purposes. The objective should not be to eliminate those boundaries indiscriminately, but to make them traversable when a decision crosses them.

Flattening those differences into one schema may make information easier to count while making it harder to trust. Which source remains authoritative? When was a relationship observed? Was it imported, inferred or confirmed? Who is responsible for correcting it?

These questions are not metadata overhead. They determine whether the information can support a consequential change.

Context must remain connected to provenance

A useful architecture environment should make information easier to combine without pretending that every source has become identical.

That means preserving provenance as knowledge is accessed and connected. An architect should be able to distinguish three different kinds of knowledge:

Observed, imported, inferred and confirmed are not interchangeable states. Traceable reasoning depends on keeping the distinction visible.

This approach also changes migration strategy. Instead of forcing every source into the destination before value can appear, an organization can expose a bounded source, retain its structure and connect it to another source for a particular decision. The result can be evaluated before the next connection is made.

The operating model matters more than the container

Repositories remain valuable. Reliable platforms, backups, access and support are fundamental. But the platform is one part of the capability.

The larger question is how knowledge moves from capture to use. Who maintains it? How is it related to operational evidence? What happens when a platform is retired? How can a decision-maker distinguish confirmed context from an inference?

The operating model therefore needs to preserve authority at the source while making selected knowledge usable across source boundaries. ARC does not require every relevant fact to be copied into one repository. It requires a controlled way to discover, relate and inspect the context needed for a change.

The goal is not to abolish repositories. It is to stop asking a repository to solve a problem that exists between systems, teams and responsibilities.

A source of truth is valuable. A controlled way to relate several sources of truth is an execution capability.

Federate authority. Connect context.