Two biases.
One conversation.

04 / Put it into practice

Ask for challenge. Build in verification.

Give the model the observed problem, intended outcome, relevant evidence and constraints. Then try this:

Before proposing solutions, separate the observations from my assumptions. Where might my wording, selected evidence or requested role steer the answer? Offer credible alternative explanations and identify whose perspective is missing. What evidence would distinguish between those explanations? What would change the recommendation? Identify uncertainties, link material claims to checkable sources, and suggest one small test with the people affected.

This can expose blind spots. It cannot certify an answer as unbiased. Check the sources yourself. Different perspectives deserve examination in proportion to their evidence; inventing an equally strong “other side” is not balance.

1. Set the criteria first. Agree what a good outcome means before evaluating options. Assess each criterion separately.
2. Change the frame. Start a fresh conversation without the preferred conclusion. Try another model or evidence set and investigate differences. Agreement is not independent proof.
3. Go back to people and evidence. Speak to affected customers and colleagues. Test assumptions against observations, not invented quotes or personas.
4. Check the pattern over time. For repeated use, test relevant groups and realistic cases, record failures and revisit checks when models or workflows change. Keep an accountable person at consequential decisions.

05 / Faster answers. Stronger understanding.

Read it. Understand it. Own the decision.

AI can help us reach a useful answer faster. Producing that answer and understanding why it is right are different achievements. If we cannot explain the reasoning, spot missing context or recognise when the situation changes, a polished output can conceal a gap in our capability.

The emerging evidence concerns cognitive offloading, learning and independent judgement. It does not establish that ordinary AI use causes general clinical cognitive decline. Some studies find weaker learning when people delegate the thinking; others show that AI and explicit reasoning practice can complement each other. The task, the design of the assistance and how people use it matter.

The widely discussed MIT essay-writing study also reported differences in recall and EEG measures. It is a small preprint. Its findings should not be translated into claims that AI “damages the brain”.^[13]

Make understanding part of the work.

TIP’s practical response starts in PERSONAL and carries through the pathway. We ask for evidence of both a useful outcome and the person’s ability to understand and challenge it:

1. Read the important material. Open original sources and check the context behind consequential claims. A summary can help you navigate; it does not establish that the source supports the answer.
2. Explain the reasoning yourself. Close the answer and describe the recommendation, assumptions and trade-offs in your own words. Revisit anything you cannot explain.
3. Ask for challenge and teaching. Explore a competing explanation, ask why a step follows, and identify what evidence would change the answer. Treat those suggestions as questions to test.
4. Apply and verify. Try the method on a different case, check the result with the person who uses it, and review what actually happened. Measure effort and outcome separately.

These are practice principles informed by the evidence, not a clinically validated prevention programme. Effectiveness means better decisions and outcomes, supported by people who can still exercise their own judgement.

06 / Make the difficult conversation easier

Challenge the assumptions.
Keep ownership with leaders.

A project can carry someone’s reputation, years of effort or a promise they have made. That investment deserves acknowledgement. The decision still needs to consider the value, costs, risks and commitments ahead. A useful question is: “Knowing what we know now, what would we choose to do next?” Stopping also has consequences; include those alongside the options to continue or change course.

Different views are not always evidence of bias. A leader may be responding sensibly to a target that conflicts with another function’s target. A change that improves one P&L may move cost or risk elsewhere. Making that tension visible allows the executive team to discuss what is best for the whole business and whether incentives need to change.

AI can help structure approved evidence, compare options, identify missing perspectives and suggest questions about possible bias. It cannot establish a colleague’s motives or certify that a judgement is unbiased. Check interpretations with the people concerned. Use the output to prepare a respectful conversation, with genuine room for disagreement.

Using only the evidence provided, separate facts, assumptions and unresolved questions. Compare the options against our agreed customer and enterprise outcomes. Where could prior commitments or conflicting incentives affect the choice? Who benefits and who bears the cost or risk? Do not infer anyone’s motives. Suggest neutral questions to test these possibilities with the leaders involved, and identify the decision that needs an accountable owner.

Transformation enables the decision. It can bring evidence, options and dependencies together, facilitate challenge and follow through on commitments. Accountable leaders make the choices within their authority. The executive team and CEO resolve enterprise trade-offs under agreed decision rights. Clear sponsorship gives the Transformation team the authority and support to deliver what has been decided.

That is the constructive leadership opportunity: make it safe to revise a view, make incentives discussable, and make responsibility explicit. Better decision support helps leaders exercise judgement and own the consequences.

Build the habit through the work.

This is why TIP combines mindset and method. The adaptive work is holding up the mirror and being willing to change our view. The practical work is reframing, asking better questions, testing assumptions and validating with real users.

Practise that in PERSONAL and TEAM. Deepen customer discovery and workflow experiments in ACCELERATE. Carry the checks into the applications and operating decisions that follow. The aim is better judgement you can repeat.