Scheduled Front Inbox Duplicate Sweep with AI Grouping

Real-time merging catches new arrivals but leaves a backlog of old duplicates. This workflow runs on a schedule, pulls every open Front conversation, uses an LLM to group threads by customer and underlying issue, merges each cluster into one canonical thread, and posts a summary of what it cleaned up.

Use it as a nightly or hourly housekeeping sweep to keep a busy shared inbox tidy, or to clean up an existing pile of duplicates before turning on real-time collapsing. Complements the inbound-trigger collapser rather than replacing it.

1. 1A schedule trigger fires (for example, every night at 2am).
2. 2The flow fetches all open Front conversations across the target inboxes.
3. 3An OpenAI step clusters threads by customer identity and shared issue, returning merge groups with confidence scores.
4. 4A logic step keeps only clusters above the confidence threshold and discards ambiguous ones for human review.
5. 5Front merges each approved cluster into its oldest thread and tags it "swept".
6. 6A Slack digest reports how many threads were merged and lists any clusters skipped for review.