MEET PILOT · CATEGORY · TYPE
Meet Pilot
Meet Pilot is a concept product for people who live in sprints and spend a large share of their week in meetings. It listens to the call, writes the summary, and turns agreed action items into real tickets in Linear or Jira, with a review step before anything gets created.
- Engagement
- Self-directed concept project
- Duration
- 1 Week
- Scope
- Product concept · web app UX · agentic conversation design
- Year
- 2025
There's a shift happening inside tech companies right now: instead of buying another generic SaaS tool, teams are building small, workflow-specific internal apps for the exact thing they do every day, and getting far better results than a general-purpose tool gives them.
Meeting notes are a good example of a workflow stuck on the generic version. Most tools listen and hand you a transcript and a summary afterward. What happens next, someone rereading the notes, deciding what's actually a task, and manually opening Linear or Jira to create it, is still on a person. Usually the product manager, after every meeting, for every action item.
Meet Pilot is a concept product built to test the workflow-specific version of this. It listens to the meeting, gives you the summary and key takeaways like any notetaker would, but it also turns agreed action items into real tickets, with a review step before anything gets created, easing the exact step that currently falls on one person's plate after every call.
This is self-directed work, not a client engagement, the same as FarmOS.
We looked at how existing AI meeting tools handle this, Fathom being the closest comparison we studied, and spoke with three to four people in corporate roles who work in sprints, live inside a ticketing tool day to day, and spend a large share of their week in meetings.
Informal conversations, not a formal research study. Flag if Fathom isn't the tool you meant.
Summaries stop short of the work
Every tool we looked at stops at the summary. Action items get listed, but turning them into tickets is still a manual, separate step, and that's exactly where things get dropped.
Nobody wants tickets created silently
The idea of an AI opening Linear or Jira on its own, without a look-first moment, made people more nervous than the idea of it missing something.
Nobody re-reads the transcript
People want a summary they can act on in ten seconds, with the transcript and recording available to check, not the default view.
The agent stopped being a feature bolted onto the dashboard and became the thing the dashboard was built around.
The early instinct was familiar: build the meeting dashboard, then attach an AI assistant to it as a chat panel on the side. We moved away from that. If the agent was going to be the reason this product existed, treating it as an add-on undersold the whole idea.
So every screen got optimized around the agent's actions being visible where the work already lives, not just inside a chat thread. Ticket creation shows up directly in the Tickets Created panel with its own approve or reject state. A reschedule shows up as a running checklist on the meeting screen itself. The user isn't reading a chat transcript to find out what the AI did, they're seeing it happen in the same place they'd look anyway.
This is concept work, so the next step is testing the parts that are easiest to get wrong in practice: whether people actually want a review step on every ticket, or whether that becomes its own chore once the novelty wears off, how Linear and Jira behave when a ticket's fields or labels don't map cleanly to what MeetPilot expects, and whether the notetaker needs to join calls live or whether async upload covers how the people we spoke with actually meet. There's also a stray reference to Salesforce in one of the completed-action screens that doesn't match anything in the reschedule flow, worth a second look before this ships.
Built for one meeting. Not tested on a real sprint yet.