PRD-STD-016: Channel-Specific AI Governance
Standard ID: PRD-STD-016 Version: 1.0 Status: Active Compliance Level: Level 2 (Managed) Effective Date: 2026-02-22 Last Reviewed: 2026-02-22
This page is the normative source of requirements for this control area. Use it to define policy, evidence expectations, and audit/compliance criteria.
For implementation and rollout support:
- Execution plan: Apply-Ready Rollout Kit
- Adoption sequencing: Production Rollout Paths
- Hands-on tutorials: Production Tutorials & Starter Guides
- Runnable repos / apply paths: Reference Implementations
Use the Compliance Level metadata on this page to sequence adoption with other PRD-STDs.
1. Purpose
This standard defines mandatory governance controls for AI products that operate across multiple delivery channels. AI behavior that is safe and effective in a web interface may violate platform policies on messaging channels, exceed latency budgets on voice channels, or produce unsafe outputs in constrained mobile UIs. Without explicit channel governance, organizations risk inconsistent user experiences, platform policy violations, and channel-specific safety failures that go undetected by aggregate monitoring.
2. Scope
This standard applies to:
- Any AI product feature deployed across more than one delivery channel
- Any AI feature deployed on a third-party platform with its own terms of service and content policies
- Channels include: web applications, mobile applications, REST/GraphQL APIs, messaging platform integrations (WhatsApp Business, Telegram, Slack, Microsoft Teams), voice interfaces (IVR, voice assistants), email, and embedded widgets
This standard does not replace PRD-STD-009 through PRD-STD-012. It adds channel-specific controls required for multi-channel AI product operation.
3. Definitions
| Term | Definition |
|---|---|
| Delivery Channel | A distinct surface through which AI product behavior reaches end users — including web, mobile, API, messaging, voice, and email |
| Channel Inventory | A documented registry of all channels through which AI features are deployed, with per-channel risk tier, owner, and compliance status |
| Platform Compliance Overlay | Additional governance controls required by a third-party platform's terms of service, content policies, or technical constraints |
| Channel-Specific SLO | Service-level objectives tailored to a specific channel's latency, throughput, and reliability expectations |
| Human Handoff | The transfer of an AI-managed interaction to a human agent, with channel-appropriate mechanisms and context preservation |
| Channel Fallback | A predefined degraded behavior mode when AI capabilities are unavailable or unsafe on a specific channel |
| Multi-Channel Consistency | The principle that AI product behavior should produce equivalent quality and safety outcomes across channels, adapted for each channel's constraints |
4. Requirements
4.1 Channel Governance Framework
REQ-016-01: Every AI product MUST maintain a Channel Inventory documenting all channels through which AI features are deployed, including channel type, risk tier, named channel owner, and compliance status.
REQ-016-02: Each AI feature MUST be individually assessed for channel suitability before deployment on a new channel. A feature approved for web deployment MUST NOT be assumed safe for messaging or voice channels without channel-specific evaluation.
REQ-016-03: Every channel MUST have a named owner accountable for channel-specific safety, compliance, and performance.
REQ-016-04: Channel governance reviews MUST occur before adding AI features to new channels and at minimum semi-annually for existing channel deployments.
4.2 Channel-Specific Safety Policies & SLOs
REQ-016-05: Safety policy boundaries MUST be defined per channel, accounting for channel-specific risks (e.g., messaging channels have higher abuse potential; voice channels have limited output correction options).
REQ-016-06: AI feature SLOs MUST be defined per channel, reflecting channel-appropriate latency, throughput, and availability targets. Voice channels MUST define sub-second response time SLOs. Async messaging channels MAY define relaxed latency SLOs.
REQ-016-07: AI-generated response format, length, and structure MUST be validated against channel constraints (e.g., character limits on messaging platforms, audio length limits on voice, screen size on mobile).
REQ-016-08: Organizations SHOULD implement channel-specific content filtering levels where platform policies or user expectations differ by channel.
4.3 Platform Compliance Overlays
REQ-016-09: When deploying AI features on third-party platforms, the organization MUST identify, document, and comply with all platform-specific content policies, messaging policies, and technical requirements.
REQ-016-10: Platform policy changes MUST be monitored and assessed for impact on AI feature compliance. A process for policy change detection and response MUST be documented.
REQ-016-11: AI features on platforms requiring template approval (e.g., WhatsApp Business message templates) MUST have template governance workflows ensuring templates are approved before deployment and re-validated after AI behavior changes.
REQ-016-12: Organizations SHOULD maintain a platform compliance matrix mapping each third-party platform's requirements to AEEF controls and identifying supplementary controls needed.
4.4 Channel-Appropriate Fallback Behavior
REQ-016-13: Every AI feature MUST define channel-specific fallback behavior for when AI capabilities are degraded or unavailable.
REQ-016-14: Messaging and voice channels MUST implement human handoff capabilities with context preservation. The handoff MUST transfer conversation history, user intent, and AI confidence signals to the human agent.
REQ-016-15: Organizations SHOULD define per-channel escalation thresholds based on AI confidence scores, safety signals, and user sentiment to trigger human handoff automatically.
4.5 Multi-Channel Consistency
REQ-016-16: AI product behavior MUST produce functionally equivalent outcomes across supported channels. Quality and safety parity gaps between channels MUST be documented and tracked.
REQ-016-17: Release testing MUST include channel-specific test cases for each supported channel. End-to-end testing MUST validate AI behavior through the actual channel delivery path, not only via internal APIs.
REQ-016-18: Organizations SHOULD implement unified identity and conversation context across channels so that users transitioning between channels experience continuity.
5. Implementation Guidance
Minimum Channel Governance Pack
Teams SHOULD establish:
- Channel Inventory template with risk tier and compliance status
- Per-channel safety policy specification
- Platform compliance matrix for active third-party platforms
- Channel-specific SLO definitions
- Human handoff protocol per channel
- Channel fallback runbooks
- Multi-channel end-to-end test suite
Example Channel Inventory
| Channel | Type | Risk Tier | AI Features | Platform Compliance | Owner | Status |
|---|---|---|---|---|---|---|
| Web Chat | Web | Tier 2 | Intent classification, response generation | N/A (own platform) | @web-team-lead | Active |
| WhatsApp Business | Messaging | Tier 2 | Response generation, appointment booking | WhatsApp Business Policy, template approval | @messaging-lead | Active |
| Mobile App | Mobile | Tier 2 | Response generation, push notifications | App Store / Google Play policies | @mobile-lead | Active |
| Voice IVR | Voice | Tier 3 | Intent classification, call routing | Telecom regulations | @voice-lead | Active |
| REST API | API | Tier 1 | All features (consumer-managed UI) | API Terms of Service | @api-lead | Active |
| Slack Integration | Messaging | Tier 1 | FAQ, status queries | Slack App Directory Policy | @integrations-lead | Beta |
Minimum Operational Metrics
Track at least:
- per-channel availability and latency against SLOs
- per-channel safety violation rate
- platform compliance audit pass rate
- human handoff rate by channel
- cross-channel quality parity gap
- channel-specific user satisfaction scores
6. Exceptions & Waiver Process
Waivers are limited to non-safety procedural controls and MUST include:
- business justification
- compensating controls
- named approver
- expiration date (maximum 30 days)
No waivers are permitted for:
- deploying AI features on a channel without channel-specific safety evaluation
- missing human handoff capability on messaging or voice channels
- non-compliance with documented third-party platform policies
7. Related Standards
- PRD-STD-009: Autonomous & Multi-Agent Governance
- PRD-STD-010: AI Product Safety & Trust Controls
- PRD-STD-012: Inference Reliability & Cost Controls
- PRD-STD-013: Multi-Tenant AI Governance
- Incident Response
- AI Product Lifecycle
8. Revision History
| Version | Date | Author | Changes |
|---|---|---|---|
| 1.0 | 2026-02-22 | AEEF Standards Committee | Initial release |