Frameworks

The Amplification Frameworks.

Original models for leading in the age of AI.

AI was built to amplify human capability, not replace it, but most organisations are managing it as if the opposite were true. These are the frameworks I use to help leaders get it right: to see where their people's value is actually moving, to understand what it really costs to trust a machine, and to know when to let it act. Each one takes a vague anxiety about AI and turns it into something you can measure, decide, and manage.

01

The Cognitive Waterline

AI doesn't take your job. It raises the waterline. Your only move is up.

Every time AI improves, it submerges another layer of human work, first recall, now production, and synthesis next. Value doesn't disappear; it migrates upward, toward judgement and intent. The Cognitive Waterline maps where your people's effort sits today and turns it into a single score for how much of it is above the line, so “upskill for AI” stops being a slogan and becomes a number you can move.

Intentask the right questionJudgementdecide, evaluate, tasteSynthesisconnect & structureProductiondraft, generate, buildRecallretrieve & rememberYour value must climbHigh groundamplified by AIAI waterlinerising every yearSubmergedcommoditised work
Amplified Submerged
Above the line54%
02

The Verification Tax

AI's real cost isn't compute. It's the work of checking what it produces.

AI doesn't remove cognitive work; it changes its shape. The effort of producing collapses, and the effort of verifying balloons to take its place, often to half of what's left. That tax is invisible, almost never budgeted, and the people best equipped to pay it are the very experts being cut. The Verification Tax quantifies what trusting AI actually costs your organisation, and shows you how to bring the rate down.

Before AI
Produce
With AIverify ≈ half the work
Produce
Verify
Produce Verify (the tax)
03

The Reversibility Dial

Delegate to AI by what you can undo, not by what's at stake.

The hardest question in AI adoption is when to let it act on its own. The Reversibility Dial answers it with two inputs, how reversible a decision is, and how high the stakes are, and one counterintuitive insight: reversibility usually matters more. It's the whiteboard tool that turns “should we automate this?” from a standing debate into a thirty-second call.

Human decidesAI informs onlye.g. hiring · legalAI actsyou review aftere.g. draft repliesAI proposesyou confirm firste.g. delete · publishAI decidesfull autonomye.g. tag · sort · routeHigh stakesLow stakesIrreversibleReversibleAI autonomy increases →Human-ledAI-led

Reversibility weighs more than the stakes