Summary

Sam asked whether polished RingSense summaries and transcripts could also prompt follow-up actions. Yes, there is a strong use case, but the safe design is to create reviewable action signals first rather than auto-changing case records from AI summaries.

Recommendations

  1. Build an Action Queue from RingSense notes. Parse summary and next-steps text into suggested follow-ups, then create reviewable tasks or queue items tagged by category.
  2. Start with low-risk categories: settlement offer mentioned, wage loss/support documentation requested, medical bill/PIP issue, client callback promised, adjuster follow-up requested, address/check mailing logistics, investigation follow-up, referral-out discussion.
  3. Add confidence and source snippets to every suggested action so staff can see why it was created.
  4. Keep destructive or case-changing actions behind human review. Do not auto-update settlement, demands, referrals, disbursements, or case status from a transcript alone.
  5. Add QA fields or a dashboard for RingSense output quality: formatted note present, transcript present, staff populated, linked to Matter or Intake, duplicate external id avoided, and next-step extraction status.
  6. Keep SMS as a separate health lane. Prior memory shows SMS can be broken even while RingSense notes look clean, so future communications QA should compare Tasks, Notes, and structured SMS independently.

Suggested implementation path

Phase 1: Reporting only. Add dashboards/list views for post-call notes with extracted tags and missing-link/format issues.

Phase 2: Reviewable task suggestions. Add Apex or scheduled job to create pending review tasks from high-confidence next steps.

Phase 3: Human-approved automation. Allow staff to approve suggested tasks or convert them into Litify tasks/checklist items.

Phase 4: Narrow auto-actions. Only after QA, auto-create very low-risk reminders such as follow-up call reminders when a clear promise exists.

Handoff

The best next build is a small RingSense Action Queue design, using the existing formatter output and live SAIL_Call_Log__c / Litify Note records. Future agent should avoid treating AI transcript content as a legal or accounting source of truth; use it as a staff-efficiency signal with links back to source note/transcript.