AI send requirements
A live automatic response depends on several independent controls: active organization, applicable automation AI toggle, pipeline/stage/lead settings, eligible channel/account, available SHVYA AI credits, permitted sender and running workers. Superadmin policy can impose further limits.
Configure languages
Save allowed Bot Languages in AI Brain and write conditional language behavior in Playbook Rules if needed. Test normal customer language, mixed-language replies and unsupported languages. A language rule cannot make an unapproved claim accurate.
Model configuration
Qualification, sales-support and post-turn summary model selection may be exposed in settings, but platform-controlled model authority and overrides are Superadmin-managed. A post-turn summary does not replace the authoritative qualification rules.
Why a conversation stops
- AI disabled at the organization, pipeline, stage or lead level.
- The selected sender is not the lead's current pipeline number.
- Credits or platform health are blocking inference.
- Provider account/session is disconnected or temporarily rate limited.
- A customer message is historical rather than a new eligible inbound.
- The channel-specific free-form window is closed and no compliant template path exists.
Diagnose before changing
Check the relevant AI switches, latest inbound time, credits, connected account and actual AI Trace. A stopped chat can have a channel/worker cause even when Playbook syntax is fine.