AI Interaction Design¶

Library AreaTemplate Library

Current PageAI Interaction Design

FocusAI behavior design

UseDefine AI interactions, review, failure, and evidence

Repository Areadocs/template-library/design/

← Previous Access-Control Design

Current AI Interaction Design

Next → Error and Exception Design

Template purpose¶

Use this template to design AI interactions at the detailed design level.

If AI is not used, say so clearly. If AI is used, define input rules, output handling, human review behavior, acceptance and rejection behavior, failure behavior, evidence, escalation, and prohibitions.

Project¶

<Project name>

Document control¶

Field	Value
Artifact owner	`<owner>`
Primary reviewers	`<reviewers>`
Status	`<draft / in review / accepted / revised>`
Last updated	`<YYYY-MM-DD>`
Related Engineering Stage	`ES-105 — Detailed Design`
Project workspace target	`docs/project-workspace/design/ai_interaction_design.md`

AI use status¶

<AI is not used / AI is used as described below.>

AI interaction register¶

ID	Interaction	Purpose	Inputs	Output	Human Review	Evidence	Status
AI-001	`<interaction>`	`<purpose>`	`<inputs>`	`<output>`	`<review>`	`<evidence>`	`<planned / deferred / prohibited / accepted>`

Interaction details¶

AI-001 — `<Interaction Name>`¶

Field	Value
Purpose	`<purpose>`
Related Requirements	`<FR/NFR/C IDs>`
Related Workflow	`<WF-ID>`
Related Interface	`<IF-ID>`
Related Data	`<entities or fields>`
Human Accountability Owner	`<owner>`
Status	`<planned / accepted / deferred / prohibited>`

Input rules¶

<rule>

Restricted inputs¶

<restricted or prohibited input>

Output handling¶

<rule>

Human review behavior¶

<behavior>

Acceptance / rejection behavior¶

<behavior>

Failure behavior¶

<behavior>

Escalation behavior¶

<behavior>

Prohibitions¶

<prohibition>

Evidence produced¶

Evidence	Trigger	Storage / Review
`<prompt/output log, review record, acceptance/rejection note, escalation record, evaluation result>`	`<trigger>`	`<location/reviewer>`

Prompt and context controls¶

Control	Purpose	Evidence
`<prompt template, context limit, retrieval boundary, redaction, grounding, policy instruction>`	`<purpose>`	`<evidence>`

Output risk controls¶

Risk	Control	Verification
`<hallucination, unsafe output, automation bias, privacy leak, stale context, overconfidence, policy violation>`	`<control>`	`<test, review, evaluation, monitoring>`

User-facing AI transparency¶

Area	Design Behavior
AI disclosure	`<how users know AI is involved>`
Limitations	`<how limitations are communicated>`
Human review	`<how review is visible>`
Appeal / correction	`<how users challenge or correct output>`

Open AI interaction questions¶

ID	Question	Owner	Needed By
AIQ-001	`<question>`	`<owner>`	`<date or stage>`

Review checklist¶

[ ] AI use status is explicit.
[ ] AI interactions are identified.
[ ] Input rules and restricted inputs are defined.
[ ] Output handling and human review are defined.
[ ] Failure and escalation behavior are defined.
[ ] Prohibitions are explicit.
[ ] Evidence is defined.
[ ] Transparency and correction mechanisms are addressed where relevant.

Continue to Error and Exception Design

Define error scenarios, safe behavior, user responses, logging, escalation, and stop conditions.

Open Error and Exception Design →