Complexity by precision matrix for deciding when to use AI agents.
A 2 by 2 matrix. The vertical axis is complexity, low at the bottom and high at the top. The horizontal axis is precision required, low on the left and high on the right. The top-left cell, high complexity and lower precision, is highlighted as a good early starting point for agent projects because the output can be reviewed or refined; example chips are Lecture summaries and Weekly learning assistant. The top-right cell, high complexity and higher precision, contains Tax forms and Customer service and returns. The bottom-right cell, lower complexity and higher precision, contains Invoice extraction. The bottom-left cell, lower complexity and lower precision, is intentionally empty because the article does not place a canonical example there. Customer emails sit near the centre as a mid-complexity, mid-precision example.
high
Complexity
low
Good early starting point
High complexity / lower precision
output can be reviewed or refined
Lecture summaries
Weekly learning assistant
High complexity / higher precision
Tax forms
Customer service & returns
Lower complexity / lower precision
Lower complexity / higher precision
Invoice extraction
Customer emails
low
Precision required
high