Play free

Articles

Read about the decisions behind the AI answer.

These articles explore the questions managers face when AI advises, acts or writes for them: what to check, how much to delegate and who stays accountable.

Start with a situation you recognise, then try it in a simulation.

Part 1 · 01 to 03

The frame

What has changed now that AI gives everyone the answers.

Illustration: two managers at neighbouring desks by day, one checking a printed contract with a phone to her ear, the other typing at his laptop without looking up.

Start here · 01

Why an AI answer still needs a human decision

Two people can hand in the same AI-assisted work and reach it in very different ways. That difference is now the one that matters.

Illustration: a wide, half-empty office floor on a stormy evening, rows of empty desks and lamps, one manager small and far away at her laptop.

Illustrative scene

02

AI used to do the work. Now it does the interpreting.

AI has moved from carrying out tasks to taking part in how we read a situation. That puts it inside the decision, not before it.

Illustration: an experienced recruiter with short grey hair leaning back at her open laptop in the afternoon sun, a stack of untouched files beside it.

Illustrative scene

03

When better AI makes experienced people worse

In a field experiment, recruiters given higher-quality AI were less accurate than those given weaker AI. The reason matters for every leader rolling out AI.

Part 2 · 04 to 06

The easy fixes

Why "always check" and "make it explain" are not enough.

Illustration: two colleagues sitting back to back in a bright office, a young man leaning close to his laptop, an older woman sitting back with her arms crossed.

Illustrative scene

04

There are two ways to get AI wrong. Most advice covers only one.

People follow AI when they should not, and they ignore it when they should not. Both are failures, and a team can show both in the same week.

Illustration: a neat printed page of plain grey bars under a desk lamp in the evening, a pen hovering above it, nothing marked.

Illustrative scene

05

A confident answer is not a checked answer

What should you check before acting on an AI recommendation?

Illustration: a balding manager with a lanyard at his laptop in soft afternoon light, one hand to his face, reading.

Illustrative scene

06

Asking the AI to explain itself does not make people check it

The common fix for blind trust is transparency. The research suggests explanations can raise reliance without making it any wiser.

Part 3 · 07 to 12

The case for rehearsal

What it takes to see judgment, and to practise it before it is real.

Illustration: a leader in a white shirt at a whiteboard drawing empty boxes joined by arrows, two colleagues at the table listening.

Illustrative scene

07

What does judgment with AI mean?

Judgment with AI means deciding what to use, what to check and what to do next. Five everyday examples, what StepSim makes visible, and where the idea comes from.

Illustration: two office colleagues at lunch in a busy hawker centre, a woman in a headscarf showing her phone to a young man, plates of food on the table, ceiling fans above.

Illustrative scene

08

Everyone agrees checking AI is their job. Most do not do it.

Almost everyone says checking AI answers is their own responsibility. Most check only sometimes, or less. The gap is not knowledge. It is habit.

Illustration: a sales director in a grey suit walking out of a glass-walled meeting room, the committee still seated behind the glass, rain on the windows beyond.

Illustrative scene

09

A good decision can end badly

How do you examine the quality of a decision separately from its outcome?

Illustration: an analyst in a white blouse and violet lanyard, upright at her monitor in a glass-walled operations centre by day, colleagues at desks behind.

Illustrative scene

10

An override count tells you how often. It cannot tell you whether.

Frameworks now suggest tracking how often people overrule AI. The number is useful, but it cannot show whether each override was right.

Illustration: a manager writing by hand in a paper notebook in the morning sun, a laptop open beside it, untouched.

Illustrative scene

11

Before you hand work to an AI agent, write down its limits

What can the agent do alone, and what should bring a person back into the loop?

Illustration: a leader in weekend clothes at a small cafe table on a shophouse walkway on a Saturday morning, laptop open, coffee beside her, a shuttered window behind.

Illustrative scene

12

Practise AI decisions with feedback

At work, you rarely find out whether the AI was right. In a rehearsal with a known answer, you do. That is why practice, followed by a debrief, matters.

Put the question into practice.

The Friday file puts you in charge of an AI rollout decision, with evidence to review and a deadline to meet.

Try The Friday file