Signal Tasks - AI Meeting and Follow-up Assistant

Challenge

In most teams a meeting ends and the decisions evaporate. Someone has to remember what was agreed, work out who owns each item, write it down somewhere nobody opens, and then spend the rest of the week asking "any news on that?". Existing task tools do not solve this, because the expensive part is not tracking the work. It is capturing it in the first place and following it up without turning the manager into a full-time reminder service. It also had to live where the team already is: Telegram, with no new app to install and no training required.

Solution

I built Signal Tasks end-to-end as a product. Send the bot a voice note or a recording of a meeting and it transcribes it, writes proper minutes, and extracts the action items, and you can correct it in plain language ("give item two to Khoei", "drop the third one") before anything is sent. It resolves each owner against the people directory, asks "which Hosseini?" when a name is ambiguous, learns the nickname for next time, and only then delivers the task to that person in their own Telegram chat. From there the product is all about follow-through: an employee can report progress but never marks their own work done, the manager approves it, and the bot chases open items on a schedule, escalating and reporting instead of nagging. Around that core I added a full mini app panel with a manager dashboard, workload per person and a monthly Gantt, a role and permission system with teams and scoped visibility, a knowledge-aware chatbot, meeting requests, personal to-dos, working-hours-aware deadlines and reminders, and an SMS fallback for people who miss Telegram notifications.

Key Features

• Meeting Audio → Minutes → Action Items
• Natural-Language Correction Before Anything Is Sent
• Owner Resolution With Disambiguation & Nickname Learning
• Automatic Follow-up, Manager-Approved Completion
• Dashboard, Gantt & Per-Person Workload
• Roles, Teams and Permission-Based Access

Project Process

Results