The problem
What needed to change
I started Jottie because there wasn’t one tool in our company workspace that joined the whole process together. In my Teams workflow, transcription had to be enabled and could be forgotten. Even when it was running, the meeting owner controlled the transcript: I could attend the meeting and hear every word, then still have to ask someone else to download it. Once it arrived, I’d clean it up in ChatGPT by hand. I wanted that whole pipeline to feel like one job.
What I made
The useful bit
I’m building an app that wakes when a meeting starts and transcribes locally. It uses calendar attendees and meeting context, separates speakers’ audio and works to identify their voices, then cross-references attendees with a BambooHR directory for names and job titles. From there, the pipeline cleans up hesitations and sentence fragments, summarises the conversation, adds tags and files the notes into matching folders. Summaries use Codex CLI through the company’s existing workspace and token budget, with the aim of fitting its compliance requirements.
Outcome
What changed
It’s still in active development, but it’s looking promising. The idea goes beyond a tidy transcript: each folder also summarises its contents, building a rolling context for a subject across meetings. Notes and context can then be copied in one click or made available to ChatGPT through MCP, so the useful parts of a conversation are easier to carry into the next piece of work.
Next
Where it goes from here
Keep developing and testing the full pipeline in everyday meetings, from automatic wake-up and speaker identification to useful summaries and folders that stay organised as more notes arrive.