# Product AI Integration Scenario

Read this file for customer-facing or partner-facing AI features, embedded AI systems, recommendation engines, copilots, conversational interfaces, or automated decision-support features.

## Typical Examples

- AI chat or drafting features in a SaaS product
- Recommendation or ranking systems
- AI-generated summaries, documents, or classifications
- Decision-support tools that influence user or business outcomes

## Typical Risk Areas

- Hallucinations or misleading outputs
- User overreliance
- Inadequate disclosure that AI is involved
- Bias, unfairness, or discriminatory impact
- Product liability, consumer protection, or deceptive-practice exposure
- Model drift after deployment
- No continuous improvement

## Required Questions

- Who are the intended users and how consequential are the outputs?
- What user decisions or downstream actions may depend on the AI output?
- What data is used for inference, personalization, retrieval, or tuning?
- What user disclosures, labels, or challenge mechanisms exist?
- What human review, escalation, or override controls exist?
- What testing exists for quality, fairness, abuse resistance, and production monitoring?
- Has an AI impact assessment been completed or required?
- What technical documentation, model card, or architecture overview exists?
- Will users clearly know AI is being used?
- What disclosures, instructions for use, or confidence/explanation cues are shown to users?
- What red-team testing, incident response, and post-launch monitoring plan exist?
- What approvals, accountable owners, and escalation paths are already in place?

## First Intake Set

Use this grouped intake set first when facts are missing:
- What is the AI product feature and what user problem does it solve?
- Who are the intended users and how much may they rely on the outputs?
- What model, vendor, or AI capability is being used?
- What data types are used for inference, retrieval, tuning, or personalization?
- Will users know AI is being used and what disclosures or instructions will they see?
- What human oversight, testing, red-team, incident response, and monitoring controls exist?
- Do you have an AI impact assessment, technical architecture, disclosure copy, or testing materials?

## Review Focus

- Risk classification and prohibited-use screening
- Transparency and user-facing disclosures
- Human oversight and escalation paths
- Testing, staged rollout, monitoring, and change management
- Affected-party analysis and non-AI legal exposure