AI Meeting Management & Automation

Closed — Private source — described by architecture only2025

Tech lead · Hashlogics

Three readings

Client: Meetings end to end, across two engagements: a workspace that runs a team's meetings — scheduling, agendas, minutes and attendance — and, for Little Tree Confections, an automation that turns every recorded meeting into assigned, prioritised tasks and a CEO dashboard that is already current.
Recruiter: Tech lead on a multi-tenant meeting SaaS integrating Zoom, Google Meet and WhatsApp, owning the AI features and the Google integrations; and automation engineer on the n8n workflows that turn meeting data into structured ClickUp and Notion tasks.
Engineer: Both halves put a model where a decision is, not where the work is: the assistant decides whether a conversation needs a meeting or just an email, and the agent returns structured output rather than prose, so every downstream node parses instead of guessing.

What it does

Summaries, notes and action items produced from Zoom and Google Meet calls
An AI assistant that decides whether a conversation needs a meeting or just an email
Meetings scheduled and participants notified through WhatsApp groups and messages
No note-taking or task entry after a meeting: every action item lands in the right list with an owner, a due date and a link back to the notes
A CEO dashboard rebuilt every morning, sorted into what to do first and what to schedule

A structured-output parser on the AI agent, so the task list is data the next node can trust
Tasks fan out to department lists, the CEO master list and the project list, each back-linked to its meeting page
The daily scan finds CEO-related work by assignment, mention or unresolved comment across every space, subtasks included
Any workflow failure routes to a global error handler — silent automation failure is worse than no automation

source private — architecture only

Source is private. Happy to walk through the architecture on a call — ammadtehseenkhan@gmail.com