A forecast earns trust when the leader can understand it, question it and see its limits before deciding.
Why it matters when the plan changes
A written forecast can be accurate and still be misread. The leader may take a yellow rating as a verdict, miss the counter-evidence, or not see which part of the call rests on thin evidence. The EU AI Act sets a comparable expectation for high-risk AI systems: the people overseeing them must be able to properly understand the relevant capacities and limitations of the system they are relying on.
The tension is between speed and understanding. A walkthrough adds a step before action, and a confident leader may feel it is unnecessary. The case for it is that people use a model's output more, and more sensibly, when they can question and adjust it: Berkeley Dietvorst, Joseph Simmons and Cade Massey found people considerably more likely to use an imperfect algorithm when they could modify its forecasts.
In practice
A chief executive receives a Team Forecast rating the commercial team yellow for a move into direct sales. In the walkthrough she asks why, sees that the rating rests on split pricing authority rather than on the people, and learns that one part of the call is blind because no customer-facing data was available. She decides to fix the authority first.
Evidence
Article 14 of the EU AI Act requires high-risk AI systems to be designed so that natural persons can effectively oversee them while they are in use.
Regulation (EU) 2024/1689, the EU AI Act, Article 14: Human oversight (2024)Overseers must be enabled to properly understand the relevant capacities and limitations of a high-risk AI system.
Regulation (EU) 2024/1689, the EU AI Act, Article 14: Human oversight (2024)People were considerably more likely to use an imperfect algorithm when they could modify its forecasts.
Berkeley J. Dietvorst, Joseph P. Simmons and Cade Massey, Overcoming Algorithm Aversion, Management Science (2018)
What it cannot tell you
A walkthrough improves understanding of a forecast; it does not make the forecast more accurate, and a persuasive explanation can make a weak call feel stronger than the evidence allows. It is optional, so its absence does not mean the forecast skipped internal quality assurance.
How Atlas reads it
Leader Walkthrough is optional support alongside a forecast: a review of the forecast with Atlas. Common Ground includes the walkthrough, adds personal feedback for every participant and an Atlas-facilitated session, and requires a forecast. Internal quality assurance remains Atlas's responsibility whether or not a walkthrough is bought. The naming table also lists the same consultant debrief as The Walkthrough.
Questions
The leader who owns the decision reviews the forecast with its author: the strategic shift, where friction is expected, the evidence for and against, what is unknown, the confidence, the owner and what will test the call. The aim is that the leader can explain and challenge the judgement before acting on it.
No. It is an optional support choice alongside a forecast. The forecast's own quality assurance happens either way. What the walkthrough adds is the leader's understanding and challenge, which matters most when the decision is consequential or the call rests partly on thin evidence.
The walkthrough is a review with the accountable leader. Common Ground includes the walkthrough and adds personal feedback for every participant and a facilitated session with agreed preparation and follow-up. Common Ground requires a forecast and is not sold as a standalone facilitated conversation.
Because a decision-maker who cannot see a model's limits tends either to over-trust it or to ignore it. The EU AI Act, Regulation 2024/1689, requires overseers of high-risk AI systems to be able to properly understand their capacities and limitations, and the same logic applies to any consequential judgement built on evidence.
No. The leader decides. The walkthrough exists so that the decision is informed: the leader can question, adjust or reject the recommended move, and the reason is recorded. Berkeley Dietvorst and colleagues found in 2018 that people used an imperfect algorithm considerably more when they could modify its forecasts.