Skip to content
A
ABC at Work
The method

The ABC
Model

A way of thinking for leaders who work with AI. When a machine answers clearly and confidently, it is easy to believe it is also right. Often it is not.

AAssumptions
BBiases
CComplexity
01 · Origins

The problem is not new. Its speed and scale are.

01 · The origin

Where this came from

For twenty-seven years, I have watched capable people make careful plans that fail when they face real-world risks.

The plan was usually not the problem. What shaped the result was what people believed, what they preferred and what they left out. The ABC Model combines that experience with research into why people make mistakes.

Our attention, information and energy eventually run out. When people become tired, the leader must keep the decision process careful and thorough.

Amit Srivastava

02 · The change

What AI changed

Machines work within limits too. An AI may lack context or take the shortest route to an answer. It often uses untested guesses or follows a question that already favours one answer. It can also miss the consequences for people you did not mention.

Human mistakes usually show up in ways other people can spot. AI mistakes are harder to see, because the answer sounds calm and certain. Human beings get tired and stop; machines don’t. An AI can repeat the same mistake in a second, across a hundred decisions.

03 · Changing AI skills

How the expected AI skill changed

  1. 2023Prompt Engineering

    Write a better question. The gap was assumed to be you.

  2. 2025Context Engineering

    Give it more of the situation, so it has to guess less.

  3. 2026Agentic AI

    Let it act on its own, over time.

  4. Bigger models. Longer memory. A model of the world.

Why I think it keeps moving

The industry knows about these gaps, and it is working on them. Bigger models. Longer memory. More context. Early attempts at a model of the world. Each one gives the machine more of what it was missing. That is why the name of the skill keeps moving.

Each move closes part of the gap. None of them closes the last part. You can give a model more context and more reach, and it still has no stake in what happens to you. It does not pay for being wrong. That cost stays with the person who signs.

Amit Srivastava

The causes are different. The leader’s job is the same: check ABC before you act.

02 · The challenge

Not every hard problem is the same kind of problem.

In mathematics, in code and in formal tests, an answer can be checked quickly against a clear rule. Business decisions do not work like that. The evidence is incomplete. People react unpredictably. Conditions change. The result may not arrive for months.

A model that can solve a mathematics olympiad problem is not necessarily competent to also decide what your company should do.

An everyday example

The map is not the territory.

A map is a useful representation of reality, but it is not reality itself.

  • What the AI plan assumed
  • Reality
ABC

The straight line is what the AI plan assumes. The winding line is reality. A, B and C mark the three places they differ.

A · Assumptions

The map is flat. The hill is not.

The route assumes that a straight line means an easy journey. It cannot feel the hill or know what the brief did not describe.

B · Biases

It favoured one route and did not challenge it.

You asked for the fastest route, so speed became the priority. The slower road that remains open in winter was pushed aside.

C · Complexity

The road changes with time and people.

The bridge may close, the market may fill the street and the school run may begin. These conditions interact and keep changing.

An AI answer is also a map. It is built from your prompt, the data it can reach, and patterns it learned from the past.

Scientist and philosopher Alfred Korzybski, 1931.

03 · The Method

The ABC Model

The ABC Model is a five-step loop you can run every time.

The AI writes the first draft. You make the final decision. The three checks go in between.

  1. The AI answers

    Start with what the AI gave you. Treat it as a first draft, not as a decision.

  2. Assumptions

    Ask the AI what it had to guess. Then check which of those guesses you can actually support with facts.

  3. Biases

    Ask for the strongest case against the answer you prefer, using your own numbers. If the answer was never challenged, it was never tested.

  4. Complexity

    Add what the prompt left out: the people this affects, how long it will take, and what happens after you act.

  5. You decide

    Make the decision. Write down who is responsible and when you will check the result.

The machine supplies speed.

The leader supplies judgement.

Continue the argument

Learn the complete method in the book.

The book works through the method with real workplace examples, exercises you can use, and a clear account of where ABC does not help.

Or start now, free

Run the checks on a decision you need to make.

The ABC tool asks you a few simple questions, then writes a prompt that tells the AI to argue against your plan.

Everything stays in your browser. Nothing is sent to a server.