One idea, five minutes

Stop optimising the horse. Decide what the car is.

A century ago the best-run stables responded to the motor car by improving their horses. Every AI initiative your firm is currently weighing sits on one side of that line: it either makes the current model run faster, or it asks what the firm should become. AI strategy is shareholder defence, not workflow speed — the question is not “where can AI save hours?” but “where will this firm’s value live when cognition is cheap?”

AI does not mainly threaten firms by doing their current work more cheaply. It threatens the assumptions that made that work valuable.

Old world to new: polishing the carriage versus constructing the car
Horse optimisation, in the wild

Sixty Post-its is not a strategy.

The morning after the AI ideas workshop: a wall of sticky notes, an empty room

Two workshops, same room, same executives. “How can AI help our business?” produced a sixty-item use-case backlog; a structural question produced three capital decisions. The difference was the shape of the question.

An AI ideas workshop
→ produces a backlog
A copilot rollout
→ produces the same firm, slightly faster
A pilot chatbot
→ produces a demo

All worth doing — as operational hygiene, from operational budgets. None of it decides what the firm becomes. That is polishing the carriage.

Deciding what the car is

Start with an asset map the board pack doesn’t have.

Under improving AI, everything your firm owns behaves as one of three classes:

Stranded
Yields from a scarcity AI is dissolving

First-pass analysis. Templated reporting. Junior-heavy delivery pyramids. Looks fine on this quarter’s P&L while the multiple quietly decays.

Convertible
Holds information AI can now read

Fifteen years of proposals, project files, client email and delivery history. Low-yield today; transformational once compiled — the asset class most firms are sitting on and not counting.

Compounding
Appreciates as AI improves

Evidence trails, test harnesses, governed delivery machinery, compiled domain context. Looks like overhead in the old world; becomes the moat in the new one.

A typical mid-market firm today allocates roughly 90% stranded · 8% convertible · 2% compounding — and the board pack cannot see it, because the accounting taxonomy predates the question.

The method behind this page is published — the Terminal Value Doctrine: select AI moves by whether they defend or increase what the firm is worth under cheap cognition, not by near-term workflow ROI.

Stranded, convertible, compounding — the three asset classes

Running this map on a firm’s own evidence is what the Future Value Review does.

The review →