An enterprise architecture function has produced extensive documentation over three years and business leaders question its value. What is the function doing wrong?

What it is doing wrong

Producing artefacts rather than decisions. An EA function whose output is models, catalogues and standards documents has confused the means for the end. The models are inputs to decisions; if no decision depends on them, they are cost.

Three specific failures:

1. Describing the estate rather than changing it. A complete application inventory, maintained laboriously, that nobody consults when making an investment decision.

2. Operating at the wrong end of the lifecycle. Reviewing designs at the end, when changing them is expensive, so the function can only object. Its experience is worth most at the sketch stage.

3. No measurable outcome. The function cannot say what it changed — which duplication it removed, which investment it redirected, which risk it retired.

What a valuable function does

It answers investment questions the business is actually asking. Should we consolidate order management across regions? What does separating this business line cost? Which of these forty applications can be retired and what does that save? Which capability gap is blocking the market entry we announced?

Each of those is a decision with money attached, and each requires exactly the models EA maintains — but the models are built for the question, at the depth the question needs, rather than uniformly.

The practical reorientation

• Asymmetric depth. Two levels of capability decomposition everywhere, five where a decision is pending. Uniform depth is what makes EA expensive and unused.
• Overlays that produce decisions — strategic importance against current maturity, cost against usage, duplication counts. The value is in the intersections, not the structure.
• Engage early and advisorily, with a mandatory gate reserved for a narrow set of genuinely significant triggers.
• Provide, do not only require — reference architectures, worked examples, a paved road. A function that only asks questions is a tax.
• Measure outcomes: applications retired, duplicate capabilities consolidated, standards adopted voluntarily, decisions influenced with the money attached.

The test

Do teams and business leaders seek EA out before they are required to? That is the only reliable indicator that the function delivers value rather than extracting compliance — and if the answer is no, more documentation will not change it.