Approach

How I think about AI in the enterprise.

Not a service menu. A territory, the questions I keep returning to when AI ambition meets the reality of a regulated, complex organisation. Most enterprise AI doesn't fail on the model. It fails on the operating model. The work is building the system underneath. I call it the Living Stack: a living knowledge substrate, the four layers that turn it into governed, coordinated, running work and accountable judgment, and the human uptake that has to run through all of it. AI doesn't create value. Systems do. And AI value is only real when it moves the bottom line.

The Living Stack

Apex

Impact

The measurable change the system makes, net of cost: what you save, what you speed up, and what the output changes. All of it converges on the bottom line.

Activity is not impact.

Running through all four

Adoption

Whether people actually use what was built, across every layer, not whether they were trained once.

A tool deployed is not a tool used.

The four layers

  1. Layer 4

    Judgment / Decision

    Where AI assists, where it decides, where it stays out. The accountable scrutiny that catches a confident wrong output before it ships.

    AI exposes judgment.

  2. Layer 3

    Execution / Workflow

    AI embedded in the systems people already work in, not demonstrated in a sandbox. Running, not pitched.

    If it doesn't run, it doesn't matter.

  3. Layer 2

    Coordination / Alignment

    Strategy, technology, legal, medical, commercial, all using the same words, the same definition of value, the same view of risk.

    AI fails between teams.

  4. Layer 1

    Governance / Trust

    AI governance designed to accelerate good decisions, not slow every decision down. The frame that makes speed possible.

    Trust is designed, not audited.

Substrate

Knowledge

What the organisation knows, kept alive: claims with provenance, not documents in folders. The ground the whole stack stands on.

Content is frozen knowledge. AI runs on what you've kept alive.