Map
Framework

The AI delegate matrix

The question it answers

When a team lead hands out work to an AI-assisted team, which tasks can AI do alone, which need AI to draft and a person to review, and which stay human, given the stake and the ambiguity?

Origin
Dell'Acqua et al., 2023; Parasuraman et al., 2000
Diagram type
2×2 matrix
Level of analysis
Individual
Topics
AI for leadersStrategy & leadership
Glass sphere with a glowing neural-network core, icon for AI for leaders

Diagram

The AI delegate matrix

LowAmbiguity of the taskHigh
AI drafts, human reviewsMedium risk
Human onlyHigh risk
AI does it allLow risk
Human does it, AI challengesMedium risk
LowStake if wrongHigh

When to use it, and when not to

Use it when

Use it when you list the week’s tasks and must choose which to give to AI. Score each task on two axes: how much damage a mistake causes, and whether the brief is clear or ambiguous.

Avoid it when

The two axes are this site’s grouping of the sources above; none of them proposes this four-cell matrix. Ambiguity is only a rough indicator: if you have already tried the task with AI and can check the result, trust the trial over the cell in the matrix.

How to apply it, step by step

  1. Score the stake if it goes wrong. The cost of consequences is one criterion for choosing a level of automation ((Parasuraman et al., 2000)).
  2. Score the ambiguity. In an experiment with 758 consultants, AI made people faster and better on tasks inside its capability frontier, but on a complex managerial task outside it, people using AI were less likely to reach the correct answer than people without AI ((Dell'Acqua et al., 2023)). The frontier is not visible from how hard a task looks, so treat ambiguity as a rough estimate and run one trial.
  3. Read the cell: low stake and a clear task, AI does it all; low stake but an ambiguous task, AI drafts and a person reviews; high stake but a clear task, the person does it and AI challenges; high stake and an ambiguous task, people only.

Pitfalls

Pitfall

Putting a task in “AI does it all” just because it is easy for a person. Easy for a person is not necessarily inside AI’s frontier ((Dell'Acqua et al., 2023)).

Sources

References

Listed A–Z by author surname, Harvard style. A source with a DOI links to https://doi.org/…; one without links to its page or says plainly that it has no DOI.

  1. Dell'Acqua, F. et al. (2023) Navigating the jagged technological frontier: field experimental evidence of the effects of AI on knowledge worker productivity and quality. (Harvard Business School Working Paper No. 24-013).doi:10.2139/ssrn.4573321Cited in 2 articles
  2. Parasuraman, R., Sheridan, T.B. and Wickens, C.D. (2000) ‘A model for types and levels of human interaction with automation’, IEEE Transactions on Systems, Man, and Cybernetics, Part A: Systems and Humans, 30(3), pp. 286–297.doi:10.1109/3468.844354Cited in 2 articles