INS-0009 · M4TIC Task-Fit Boundary Map
Five different jobs hidden behind “AI”
The technology matters less than the work being done. Each lane produces a different result, demands a different control question and fails in a different way.
A chatbot, for example, may generate a response, retrieve data and call tools in the same conversation.
- 01
Generate
Expected outputA draft, summary, translation, image, code sample or transformation.
Control questionCan the result be evaluated before it creates a consequence?
Typical failurePlausible content that is false, incomplete, off-brand or too similar to existing material.
- 02
Predict or classify
Expected outputA score, probability, label or category.
Control questionAgainst which samples, thresholds and costly error types has the result been validated?
Typical failureA wrong classification that looks precise because it uses a number or decisive label.
- 03
Retrieve
Expected outputAn existing record, document, source or indexed item.
Control questionIs the source correct, current and authorized for this person and purpose?
Typical failureMissing, stale, irrelevant or wrongly sourced information.
- 04
Apply rules
Expected outputA repeatable result derived from explicit conditions.
Control questionIs the rule complete, tested, versioned and able to handle exceptions?
Typical failureIncomplete or incorrectly implemented logic—not generative variability.
- 05
Act
Expected outputA modified record, sent message, initiated payment or triggered process.
Control questionWho authorized the action, what limits apply, and how can it be stopped or reversed?
Typical failureAn unauthorized, duplicated or hard-to-reverse effect.