Run your business on meeting context

We just ran our first Customer Show & Tell, with @Christian Ulstrup, Founder & Principal of Caritas Venture Co. Here's the simple truth behind it: most teams treat a meeting as something that happens and then gets forgotten. At Caritas, every client call and team conversation lands in Fireflies and becomes shared context that AI tools pull from to build, brief, and draft across the business. This session was about seeing that up close, in Christian's real workflows.

Here's what Christian showed and where to go deeper.

Building software live on the call (7:15)

Christian's team gives Claude Code access to the Fireflies live transcript during client calls. The client talks through what looks off, and the agent listens, answers, and fixes the web app before the call ends. Christian didn't even look at the app being built: he was walking a trail, and the feedback loop ran on the conversation itself. Want to try it? Start by connecting Fireflies to the AI tool you already use.

Powerline briefings after every meeting (11:57)

The Caritas team is async and fractional, so Christian built Powerline on top of their Fireflies Enterprise account. After every meeting, including the ones he wasn't in, he gets a short Slack read with the subtext worth knowing. After his own calls, a "hot wash" points out one quick win he would have missed. And every Friday, a weekly email ties the whole team's meetings together like a note from a trusted adviser.

Onboarding from the source (16:01)

When someone joins a project, an agent gathers the meetings that matter and sends them over, with shared access that expires on a set date. New teammates read the primary sources, ask follow-up questions, or watch the tape, and they get productive much faster.

Case studies from the record (18:03)

Every engagement ends with an internal case study drafted from the transcripts, with claims footnoted back to the meetings. For an updated client case study, an agent drafted it, iterated with the sponsor over email until it was signed off, and published it, fact-checking against the transcripts the whole way. See the results in the Caritas case studies.

Probability scores that keep everyone aligned (21:10)

In Arlos, the Caritas operating system, the person driving each outcome drafts an odds-of-success score from the week's meetings, then applies their own judgment. When a score jumps from 40% to 70%, or drops from 80% to 20%, that's the signal to dig in, replicate what worked, or step in and unblock the team.

A few questions came up worth calling out. Daniel asked whether relying on AI means missing the human context, like power dynamics. Christian's take: pulling every past conversation into one place lets you see more than you could hold in your head, but it's a lens, and the judgment stays with you. On keeping a human in the loop, you can't take the human out, but you can widen the loop once you have a clear goal and good context. Sam asked why the live transcript matters, and the answer is speed: changes happen while everyone is still on the call. Arthur asked whether to start with Live Assist or a custom setup: start with whatever has the least friction, push it as far as it goes, then build further. And Christian's first tip for every client is dictation: stop typing, start talking, with Fireflies Talk.

Over to you: which conversation in your week would you put to work first? If you've been there, you know the feeling:

❌ Asking "what's the update?" about a meeting you weren't in
❌ Onboarding someone by retelling months of client history
❌ Rewriting the same project update every Friday
❌ Writing a case study from memory long after the project ended

👇 Tell us the workflow from Christian's session you'd build first. Be specific. Which meetings feed it, and where should the output land? We'll help you set it up.

Keep the conversation going: still have a question for Christian, or about connecting Fireflies to your AI tools? Drop it in the comments here or in #Ask the Community, and we'll help you put your meetings to work.