The answer-key model is a training pattern in which the employee attempts the work first, the AI grades the attempt, and a manager discusses the differences. McKinsey proposes it as a way to build skill with AI in the loop rather than to replace the practice.
The answer, in one line
A training pattern proposed by McKinsey in which the employee attempts the work first, the AI grades the attempt, and a manager discusses the differences with them.
Definition#
Answer-key model: “The employee attempts first, the AI grades the attempt, and the employee sits down with a manager, who discusses the differences.” The name and the sentence are Bryan Hancock and Charlotte Seiler’s (McKinsey Quarterly, 14 July 2026). It is a proposal, not a tested method.
Why the order is the point#
The same AI can be used in the reverse order: it drafts, and the employee reviews. In that order the employee sees the answer before making any attempt. The answer-key model puts the attempt first. McKinsey’s article describes a medical-student pilot in the same terms: “The student goes first, and AI grades the attempt.” Graded entry.
The article contrasts the attempt-then-check loop with “borrowed competence”, meaning skill that exists only while the tool is present. It uses the phrase once and does not define it, so the contrast is the article’s, and the definition of the phrase is not.
Support so far is by analogy, with no outcome data#
In the text read for this page, McKinsey reports no outcome measurement for the model itself. The article is a consultancy piece in McKinsey Quarterly, written by a partner and an associate partner at the firm, and it sits alongside survey figures about graduate pessimism and entry-level hiring that are context rather than evidence for the model.
The nearest controlled evidence points the same way without testing it. Bastani and colleagues found that students given unrestricted AI access scored 17 per cent lower than students who never had it once the tool was removed, while a tutor version that gave hints and withheld answers largely removed the harm. Graded entry. The designs differ: that tutor withheld the answer, whereas the answer-key model grades an attempt after it is made. Both put effort before assistance, which is the common thread, and the link is by analogy.
The constraint is the manager’s time#
This section is interpretation. The model keeps a person in the last step, and a manager discussing differences is the expensive part. If it is dropped, what remains is an AI marking an employee’s work with nobody checking the marking. Anyone adopting the pattern should decide in advance who holds that conversation and how often, because the saving the tool offers is the first place the pattern can be cut. For where this sits in the wider argument, see how juniors become senior.
Key sources
- Hancock, B. and Seiler, C. (2026). Building expertise in the age of AI: Who trains the next generation? McKinsey Quarterly, 14 July 2026. Graded entry.
- Bastani, H. et al. (2025). Generative AI Without Guardrails Can Harm Learning. PNAS. Graded entry.
Related SuperSkills research#
See never-skilling, retrieval practice and desirable difficulty. The model and its name are McKinsey’s, not a SuperSkills coinage.
About this research#
Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.
Explainer · SS-2026-422 · Graded against the published rubric · 1 argued perspective
Hirji, R. (2026). What is the answer-key model?. The SuperSkills evidence base, SS-2026-422. https://thesuperskills.com/research/what-is-the-answer-key-model. Last reviewed 12 October 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work