The AI‑Native Score · free · 12 minutes

Most AI programmes cannot prove they changed anything.

Not because the technology failed. Because nobody wrote down what they were changing before they started — and the counterfactual disappears as it passes.

This is the free diagnostic at the centre of The Compounding Company — 25 questions, seven dimensions, scored out of 100. It is deliberately built so that buying tools alone caps you around forty.

0102030405060708090100
What you builtCeiling
Take the score Read about the book No account. No call.
The instrument

Seven dimensions. Technology is 18 of them.

Every comparable framework inverts that. This one weights what your work costs, and whether your revenue survives, above what you have installed — because that is the order in which companies actually fail.

Each question is scored 0 to 4 against fixed anchors. Two people assessing the same company honestly should land within a few points of each other.

D1Knowledge & context layer15
D2Agent & automation surface15
D3Model & infrastructure strategy10
D4Data, integration & agent security15
D5Org & operating model10
D6Economics & outcomes17
D7Business model exposure18
Four ceilings

A ceiling caps your total regardless of everything else. Where more than one applies, the lowest governs.

55

No measured outcomes

Infrastructure with no measured change in cost or output is procurement, not transformation.

50

Undefended revenue model

Efficiency gains under effort-based pricing shrink revenue rather than margin.

45

No knowledge layer

Agents without company context are demos. They do not survive contact with real work.

60

Capability past its defences

Agents acting autonomously while security is weak. Conditional — it does not apply at lower autonomy.

Free tools · CC BY-SA 4.0

Three canvases you can use, adapt and teach.

One page each. Built to be filled in by a group in a room rather than by an analyst at a desk — because the disagreement between functions is usually the finding.

The book

The Compounding Company

The instrument is chapter three. The rest of the book is what moving each score actually requires — the knowledge layer agents need and almost nobody has, the evaluation gate where the economics turn, the metrics that survive a CFO, and the question most transformation programmes never ask.

Fifty-one chapters. It weights technology capability at eighteen points out of a hundred and defends the choice. And it publishes its own weights as a hypothesis with a stated test, on the grounds that a book demanding evidence should be willing to supply some.

The AI-Native Score
The
Compounding
Company
Praveen Singh V
Research

The theory underneath it.

Six papers. The instrument's structure and the book's central arguments come from these; the calibration does not.

Learning Markets and the Dynamic Boundary of the Firm
Coordination costs are endogenous and decline with use. Why the boundary moves.
DecisionTrace and Context Graphs
Decision rationale not captured at execution time cannot be reconstructed. With proof.
Decision Native AI
Decisions as first-class objects. Facts are not signals.
WorkRank
Reputation as net marginal contribution — value created minus coordination cost imposed.
TrustRank
PageRank extended to human trust, with cognitive limits. Rate transactions, not people.
Unified Transaction Graphs
Enterprise activity as a graph of decisions, actions and outcomes.
Work with me

The calibrated version, and what follows it.

The free score is designed to score you generously — one respondent, no evidence standard. The calibrated assessment is thirty-five questions, four interviews conducted separately, evidence demanded for every high score, and a coherence matrix across five functions.

The gap between the two is typically fifteen to thirty points, and that gap is the most useful thing either number tells you.