Legacy Account Prep via AI-Powered Slack Archive Mining: 11,200 Messages, 5 Years, Briefed in 20 Minutes
Hand a new account owner a per-client prep brief distilled from the entire team chat archive, in minutes, read-only against the source system.
What was getting in the way.
When an account owner leaves, the context sitting in team chat, client preferences, past decisions, unresolved threads, becomes effectively invisible to the next owner. Reading five years of channel history by hand is a day's work per account; multiply by a dozen legacy accounts and nobody does it, so the new owner starts cold.
Context
Universal: applies to any AI agent deployment with chat-archive exports and an owner-transition event.
How the work runs.
- 01
Trigger
a batch of legacy accounts needs transfer to a new owner.
- 02
Scope the archive correctly
enumerate archive folders at the right depth: a previous sweep had reported two channels missing because it stopped one level too shallow, so the first move is to walk the whole tree.
- 03
Ingest exports already on disk
~11,200 messages, 2021→2026, from exports the team already holds, nothing fetched live, nothing written back to the source system.
- 04
Index by account
the agent attributes messages to accounts and surfaces the decisions, preferences, and open threads per account.
- 05
File per-account prep briefs
write distilled briefs into each account's record of truth, not into the source chat.
- 06
Outcome
the transcript set plus a state record is the proof, not a job exit code.
Evidence from the workflow.
Each system has a role.
Intake
Team chat archive (local exports)
Action (briefs filed here)
Per-account record of truth
Record of truth (job provenance)
Standing state record
Why this is Coworker.
Takes assignments and reports back. You hand it work, it prepares and returns a result.
- 01
Read-only against local archive exports; zero outbound to the source chat system. Proof is the per-account transcript set plus a state record, not a job exit code. Standing procedure now enumerates archive folders to full depth after one scan missed two channels for the same reason.
Impact / Outcomes
~11,200 messages spanning five years ingested from local exports in roughly 20 minutes; nothing written back to the source chat system.
Two channels a prior sweep had confidently reported missing were recovered: the earlier scan had enumerated archive folders one level too shallow.
Each legacy account received a distilled prep brief instead of raw transcripts, so the new owner starts hot.
Find where a workflow like this fits.
Start with the systems, work, constraints, and authority already present in your operation.