Skip to main content

Multi-WABA — AI Agent Impact

The Platform/WhatsApp/Billing initiative "1 CID Multiple WABA" (Lingga Mahesa, Confluence QON/51170246951) moves Qontak from 1 company = 1 WABA to 1 company = N WABAs, and scopes WhatsApp templates and broadcasts per waba_id — cross-WABA template sends are rejected with 422 — Template not valid for this WABA. Target GA: September 2026 (Meta Coex GA).

This initiative folder is the chatbot/AI-Agent dependency & impact assessment of that external change — what stays stable, what the source PRD under-states, and the dependencies the AI Agent owner must register before Multi-WABA GA. Full analysis: rfcs/ai-agent-impact.md.

TL;DR​

The AI Agent is mostly insulated — routing, KB/RAG, actions, and agent identity key off channel_integration_id / organization_id, never waba_id. The source PRD's "Chatbot / Automation" edge cases are accurate. But three items fall to the AI Agent owner:

  1. Bot-sent WhatsApp templates/Flows have no waba_id → latent 422 for multi-WABA orgs.
  2. AI Agent ↔ channel binding is single-channel and auto-picks the first channel → brand-blind for multi-brand orgs; needs a design decision.
  3. AI Agent telemetry (activity-log datamart + scorecard) has no WABA/brand dimension → can't report AI performance per brand for exactly the enterprise clients this unlocks.

Scope Changes​

  • Backend — waba_id awareness for bot-initiated template/Flow sends.
  • Frontend — AI Agent ↔ channel binding for multi-WABA orgs.
  • Data — waba_id/fbbm_id dimension on AI activity-log datamart + scorecard.

QA Lane​

Lane B — keeps a human QA gate. Out-of-reach: depends on the external Meta WABA platform that automation can't fully drive. No E2E test specs exist for this initiative yet, so the Lane-A entry bar (100% E2E, spec-mapped coverage) is unmet regardless. Classified 2026-06-29.