Turning a two-minute voice note into work a fund can act on

Olma started as an internal tool. We dictate constantly, we were paying for Wispr Flow to do it, and we wanted something narrower: press a key, speak, and have the text land where the cursor already is. So we built our own.

A fund we work with had a different problem that the same idea solved. They wanted to hear from every company they had backed, openly and often, without asking anyone to fill in another form. We rebuilt the tool around that, and what follows is what we delivered for them.

What the fund was already good at

The fund helped its companies in fairly specific ways: an introduction to the right person, an answer someone else had already paid to learn, a piece of technology another portfolio company had built. Partners did this instinctively for the founders they spoke to most.

The limit was coverage rather than willingness. Nobody holds a live picture of thirty companies at once, so help flowed towards whoever had been in the room most recently. What was missing was a low-effort way for every company to say what was on its mind, early enough for the fund to do something useful with it.

Two minutes, on a phone

There was no prompt and no reporting schedule. A founder picked up their phone whenever something was on their mind, opened the app we had built for them, and talked for a couple of minutes.

Speaking did real work here. A written update is a document, and documents get edited until they sound composed. Speech runs closer to thought, so what founders recorded tended to be whatever was actually on their mind: the hire they were unsure about, the customer conversation that stalled, the decision they kept moving to next week's list.

The app sorted what was said into wins, problems, things that were stuck and things they needed, and filed it under that company. Nobody filled in a form, nobody was chased for one, and nothing waited for a board pack.

It went to the fund's own servers

How openly people talk depends on where the recording goes.

Everything landed on the fund's own servers rather than ours or anyone else's. A founder describing a difficult quarter was talking to their investor, not to a supplier, and the fund kept what was said under the same rules as everything else confidential it holds. That mattered more than any feature in the app, because a founder who is unsure where a recording ends up says something safer and less useful.

From a voice note to a to-do item

Collecting the entries was the easy half. The work was turning them into something a partner could pick up on a Tuesday morning.

Each entry was read against what the fund actually had: the skills in its team, the people it knew, the contacts already in its records, and the rest of the companies it had backed. A company being put through a large customer's security checks was matched to the partner who had been through it twice. A hiring need was matched to someone the fund had already met. A technical problem showing up in three companies at once became something worth building once and sharing with all of them.

The output was not a dashboard. Each match became a to-do item in the fund's own project management system, attached to the company it was for and carrying the reason it came up. It arrived where the team already worked rather than in another tab nobody opened.

Some matches needed a person rather than a record. Those went to Slack, as a short message to the partner best placed to act, carrying the reason it had come up and a way to say yes without leaving the channel.

What it changed

A fund hearing from every company, in their own words and in their own time, stopped depending on who happened to call. Help reached the whole portfolio rather than the companies nearest a partner's inbox, and it arrived while it was still worth having.

The tool started life replacing a dictation app we were paying for. What made it valuable was never the transcription. It was the short path it created between a founder saying what they needed and somebody acting on it.