Four questions before AI.

A signal chain, not a use case.

Take any detectable event in a value stream — a fraud pattern, a churn flag, an arrears signal, a stock-level breach. The instrument walks the chain that follows it, and each read maps onto one interval of the signal-response distance the working paper measures:

• Detected — does the organisation see the event when it fires? The sensing interval (t₀→t₁), and the place AI vendors point at by default.

• AI-accelerates — is AI generating or speeding up detection of this event? Not a virtue in itself: an accelerant is only as good as the chain it pours into.

• Converts — once seen, does a committed response reliably follow? The commitment interval (t₁→t₂): decision rights and risk posture, not technology.

• Absorbs — can the organisation execute the committed response at the rate required? The absorption interval (t₂→t₃): reconfiguration capacity, the read most AI business cases never make.

Triage an event.

Start from one of the worked examples, or toggle the reads to match an event in your own organisation. Nothing is recorded — the instrument runs entirely in this page.

The differentiating output.

Four of the five verdicts route effort: instrument detection, fix commitment, fund absorption, or protect a loop that already works. The fifth is the one this instrument exists for. AI makes it worse fires when AI is accelerating a signal into a chain that cannot convert or absorb it — the model performs, the spend grows, and the distance between knowing and responding widens rather than closes.

An opportunity scan can only ever say yes in different volumes. An instrument that models conversion can say no.

That verdict is why the triage runs on the signal chain rather than the use case. Whether AI could detect something is a property of the technology. Whether detecting it earlier produces a response is a property of the organisation — its decision rights, its risk posture, its reconfiguration capacity. The six delay stocks give those properties names; this instrument reads four of them in sixty seconds per event.

From one event to a funded portfolio.

The page you are on is the single-event read. The full instrument runs the same triage across an entire value stream — stage by stage, event by event — and derives the AI-engagement candidates from the architecture: each one gated into worth pursuing, fix first, detection play, or AI makes it worse, with the why attached. A cost-of-delay model then quantifies what closing each gap is worth.

It runs in two settings: as the capstone of the AI Literacy for Performance executive training, where attendees triage their own value stream live, and inside diagnostic engagements, where it is the structured front door to the full SRD placement.