Builds

RADAR: meeting transcripts to Jira tickets

A TPM who isn't a developer built an app that turns a pasted standup transcript into suggested Jira tickets and weekly business reviews, cutting about four hours of work to minutes.

RADAR is an internal app built by Matias Singer, a technical program manager at Jordan and Sam's company, who demoed it on Episode 39. Matias isn't a developer. About two months before the episode he started building with the help of Claude and OpenAI Codex, and RADAR is the result: paste a meeting transcript, get suggested Jira tickets, review them, and create them in a couple of clicks. He said it took a job that used to take four hours down to minutes.

Why we built it

As a TPM, Matias spent his mornings in standups with engineers. If he was presenting, he couldn't type the updates. Because he isn't a developer, he sometimes misheard what someone said or how they planned to continue. Then he had to turn his notes into Jira tickets, and when a ticket wasn't raised in time, work got delayed. "These meetings were like the worst thing to have," he said.

When he started, he asked the developers what they did to automate their daily work, what was repetitive and how many tools they used. The goal was to gather all of it in one place. He also wanted reporting that was more dynamic than a doc full of words, which is where the charts came from.

The stack

  • Claude and OpenAI Codex: the AI coding help Matias used to build the app.
  • Granola: records the daily standups and produces the transcript.
  • Jira: where the approved tickets are created, and the boards RADAR mirrors.

How we built it, step by step

  1. Start from the repetitive work. Matias asked developers what their repetitive tasks were before building, then used Claude and Codex to build the app.
  2. Mirror the boards. RADAR has a board section that looks like Jira, with internal projects (social, marketing, IT) and every ticket marked open, in progress or done.
  3. Add analytics. An analytics view shows the same work as charts, so status meetings aren't spent reading a doc.
  4. Paste the transcript. After the standup, paste the Granola transcript into the transcript section. In the demo the transcript was in Spanish.
  5. Let the AI suggest tickets. RADAR picks out the action items and suggests tickets, with the project's members listed as possible assignees. The demo produced five.
  6. Review before anything is created. Open each suggestion, edit the title or assignee, select or deselect, then create the chosen tickets on the board.
  7. Generate the weekly business review. In the WBR section, upload one or more transcripts and RADAR writes the weekly report, split by project, with important matters and upcoming priorities. You can edit it afterward.

How it turned out

"You just saw, like I just pasted transcript and just populated and suggested me the tickets that we needed to create today," Matias said after the demo. Asked whether the suggestions were accurate, he answered: "Absolutely. We still need to control and be like this overhead on this stuff. But yeah, it's pretty accurate."

What started as a tool for his own work is now available to every employee, and Sam Nadler said it has expanded to multiple teams across the company. Editing in the WBR was added the day before the episode, after Sam suggested it, which shows how fast the app changes once other people use it.

What we'd do differently

The main design choice is the one we'd keep: a human stays in the loop. Sam called out that RADAR doesn't push straight to Jira, because not everything said in a meeting is a ticket. Matias also admitted the WBR "sometimes it's not perfect," surfacing things that aren't that important, which is why it's editable.

His roadmap covers the gaps he still sees:

  • Include tail tickets in the WBR report.
  • A pre-meeting section so every team sees where they left off the day before.
  • Store past transcripts so the tool can give guidance across meetings.
  • Connect Granola directly so nobody has to paste a transcript at all. In his words, soon it will be "just joining the meeting and conducting it."

The rest of the episode covers how the company got sales, marketing and finance teams building their own apps, starting with a weekly AI meeting led by a senior developer and split into technical and non-technical sessions.

FAQ

Can AI turn a meeting transcript into Jira tickets?

Yes. In RADAR you paste a standup transcript (it worked on one in Spanish), the AI suggests tickets with project members as assignees, and you edit and select which ones to create in Jira.

Does RADAR push tickets to Jira automatically?

No, on purpose. Sam Nadler pointed out that not everything said in a meeting should become a ticket, so RADAR queues suggestions and the user reviews, edits, selects or deselects them before creating anything.

How accurate are the AI-suggested tickets?

Matias Singer called it pretty accurate, but said a person still needs to stay on top of it and check the suggestions.

What tools was RADAR built with?

Matias built it with the help of Claude and OpenAI Codex. His team records standups with Granola and the tickets land in Jira.

Can AI write a weekly business review from meeting notes?

RADAR's WBR section takes one or more uploaded transcripts and generates an editable weekly report split by project, with important matters and upcoming priorities.