Pochtalyon AI

Important work was split across Slack, Google, and YouTrack. A missed reply, meeting, or task update could disappear between them, and even a collected item was useless if the user could not tell whether its source was connected or its next action would succeed.

Designing for the breaks around the summary

I followed what happens around each message, meeting, and task update. Before collection, the source must be connected. If that connection breaks, the user needs a recovery path. A digest needs a schedule, and a suggested action needs the connection required to complete it. I made these conditions part of the queue so a useful summary would not leave the user with a dead end.

Keeping the work in Slack and the state in the backend

I kept review, recovery, digests, and actions in Slack so users could handle the queue where the work already arrived. n8n routed intake, ranking, tools, and delivery; the backend owned authentication, connection state, schedules, and provider adapters. This kept editable workflow logic separate from durable user state.

The private product shipped as one recoverable queue

The private product shipped as one ranked, recoverable queue in Slack. Users could move from a priority to its available action or recovery path without searching across three tools. Usage impact was not measured.