Skip to content
Back to Blog
AI & Governance

What we're building next: findings that prove your directives

Shi YanJune 22, 20265 min read

Open the Findings page in most governance tools and you see a long, flat list of problems: a severity, a title, a project, a status. It's a defect tracker. Useful — but it quietly answers the wrong question.

In cajeX, a finding is never just "a problem in a project." It's the output of checking a project against your directives — the rules your organization has decided to govern by. The AI review reads the project material, evaluates it against each approved directive, and writes a finding when it sees evidence of conformance or violation. The directive is the question; the finding is the answer.

So the most important question a finding can answer isn't "what's wrong here?" It's "is this rule being followed?" That's the question we're putting front and centre.

The reframe in one line: stop treating a finding as a defect found in a project, and start treating it as a measurement of a directive's conformance.
Closing the directive → finding → directive loop A directive is the rule. An AI review checks a project against it and writes a finding — evidence carrying a conformance verdict. That verdict becomes the directive's conformance status, closing the loop. Directive the rule you govern by conformance status AI Review checks a project against it Finding evidence + verdict the finding's verdict becomes the directive's conformance status

Figure 1 — The rule produces the evidence; the evidence becomes the rule's conformance status.

The loop that was never closed

The data to answer that question already existed. Every finding records the directive it came from. Every AI review emits a per-directive verdict — conformant, partially conformant, non-conformant, or insufficient evidence. The pieces were all there.

They just never came back together. The verdict was trapped inside a single review run. The directive page knew nothing about the findings it had produced. And the Findings list showed you the symptom without the rule it belonged to.

Before


  • A flat list — severity, title, project, status.
  • No sign of which rule the finding violates.
  • No way to tell a systemic gap from a one-off.
rolls up to

After


  • Each finding shows its directive.
  • Each directive shows conformance + open findings by severity.
  • One click from the rule to its evidence, and back.

What's changing

We're making the directive the home of its own evidence, and making findings navigable as governance signal rather than tickets.

On every finding

The directive it maps to — shown in the list and on the detail, and one click away from opening that directive.

On every directive

A conformance status, when it was last evaluated, its open findings by severity, the projects they touch — and a jump straight to them.

Across the portfolio

Top directives ranked not just by how many findings they have, but by whether they're actually conformant.

A directive stops being a written rule and becomes a measured control. You can finally answer, from the rule itself: Which projects violate this? How recently did we check? Is this getting better or worse?

Why we compute it live

One deliberate choice under the hood: a directive's conformance status is computed at read time, from its current findings and the latest review verdict — not stored as a number we have to keep in sync. A governance status that can drift out of date with reality is worse than no status at all. So the directive always reflects what's true right now, the moment you open it.

Where this is going

This is the foundation for the questions a head of architecture or compliance actually asks: not "how many open findings do we have," but "which of our controls are holding, which are slipping, and where." Conformance on the directive is step one. Trends over time, and conformance as a first-class lens across the whole portfolio, follow from here.

A finding was always evidence about a rule. We're just making the product believe it too.

Ready to transform your architecture governance?

cajeX brings AI-powered reviews, knowledge management, and directive lifecycle management to your enterprise architecture team.