Requirements management

Ask why a given requirement exists, and most organizations can answer for about one hop: this ticket traces to this epic. Ask for two hops — which market signal justified the epic, and which portfolio dollar is riding on it — and the trail usually goes cold. The specification becomes an island: readable to engineering, disconnected from the reasons it was written and the money depending on it.

This dimension shares its origin with SDLC's own D4, but diverges from it in kind rather than degree. SDLC's version of this capability gates AI generation quality — can a coding agent act on this spec without a person reinterpreting it first? PDLC's version gates something else entirely: whether a portfolio investment decision has real evidence underneath it, and whether a shift in market signal reaches the specifications it should invalidate before someone builds the wrong thing anyway.

Where most organizations start (Nascent)

Specifications are written informally, with no consistent structure, and traceability back to the market or user signal that justified them is manual at best and nonexistent at worst. AI assistance in drafting is individual and unsystematic — one PM's habit, not the function's method. The first real step is mundane but load-bearing: a shared format, a minimum field set (intent, acceptance criteria, source signal), and a system of record a teammate outside the original author can actually navigate.

Where the real gains happen (Modeled → Integral)

The meaningful shift is making specifications machine-actionable — structured enough that an AI generation tool can consume one directly, without a person reformulating it first — while simultaneously building real traceability in both directions: back to the market or user signal that justified the work, and forward into the delivered artifact. Once that traceability exists, the next gain is making it bidirectional and automatic: when an upstream signal changes, the specifications it affects should surface on their own, not wait for someone to think to check.

What the top of the curve actually looks like (Telemetric)

At full maturity, the requirements layer is self-maintaining — continuously reconciling specifications against upstream market intelligence and downstream delivery signal, surfacing genuine gaps and conflicts without a scheduled review triggering the check. PM work shifts from authoring specifications from scratch to reviewing what the system has already flagged and ratifying the resolution.

Why this dimension matters

This is the bridge dimension. D1–D3 supply the signal; D8 spends the money; D4 is the only place in the model where a market read and a portfolio bet are forced to stay honest with each other, in both directions, without a person standing in the middle manually checking. When that bridge is weak, an organization can have excellent market intelligence and a disciplined portfolio process and still fund the wrong thing, because nothing forced the two to talk.

---

Drafted from the ai-native-pdlc-maturity-model's own locked v1.1.0 matrix content (2026-07-28), including the newly added per-transition verification clauses.