Tool Use

What it means

The model emits structured calls (name + arguments) that your app executes, then feeds results back into the conversation for the next step.

Why designers should care

Show tool name, scope, arguments summary, success/failure, and let users deny or undo, especially for write actions.

Example

A calendar agent proposes “Create 30-min hold tomorrow 2pm”; the UI previews event details and requires tap-to-confirm before the createCalendarEvent tool runs.

Common mistakes

• • Running write tools silently with no preview or audit trail.
• • Tool errors surfaced as generic model failures.
• • Too many tool calls in one turn with no progressive disclosure.

Related terms

• Agent: The model decides sequences of actions (search, draft, click, call APIs) within policies until the task completes or hits a human checkpoint.
• Structured Output: The product constrains generation to fields, types, and enums (often via schema validation) so downstream code and design systems can render reliably.
• Prompt Injection: Malicious or accidental instructions embedded in retrieved or pasted content override intended rules (“ignore previous instructions and…”).
• Human-in-the-Loop: The product pauses at defined checkpoints for human judgment (approve send, merge PR, publish article, charge card), even if AI drafted the action.
• Inference: Each user request triggers inference: tokens in, tokens out, optionally interleaved with tool calls and retrieval.

Explore more

• Agentic UX framework