Companies Transcribe Every Meeting. Then Dump It in a Doc Where It Dies. erra Fixes That.

The most useful thing erra's founders learned on their customer calls wasn't what companies were missing, it was what they were already doing and wasting. Teams transcribe their meetings diligently, then dump the transcripts into Google Docs where, in Kristofer's words, nothing useful happens with them. All that context, captured and then abandoned. erra exists to catch what everyone was already throwing away.

How It Started

Kristofer and Ralf had been building several products and companies at once, with different people and kept slamming into the same wall: forgetting what had been discussed, what needed doing, and what everyone else was working on. Switching between projects was painful. Giving coding agents the right context, or brainstorming with them cleanly, was just as messy.

The idea surfaced sideways. While exploring a product for event-organizing agencies, one component was a kind of company brain where every document and chat from Docs, Slack, Gmail, Notion, plus transcriptions and PDFs, vectorised so knowledge became cheap and fast to retrieve. That component quickly became the part they found most fascinating, so they cut the rest and built around it. erra brings it together in a synced application for the whole team, turning conversations into shared context that gets more valuable as more people use it.

Building From the Baltics

The Baltic languages are small markets, which means most products support them badly and that gap became erra's wedge. The team focused on excellent transcription across Baltic languages that also handles English mixed into the same conversation, which is how professional life here actually sounds.

There's a sharper insight underneath it. Many similar companies ignore the older, legacy businesses that still operate in native languages. Kristofer sees those companies as exactly the ones worth building for. Serving the customers everyone else's tooling quietly overlooks is a real, defensible way to enter a crowded category.

How It Works

erra gives both people and AI agents the context they need to work well. It connects Gmail, Slack, Google Docs, calendars, and other work tools into one continuously updated knowledge base with access governed by a company's existing permissions.

For meetings, you press record. erra transcribes and digests the conversation, adds it to the knowledge base, generates documents from templates, and pushes action items to a task board for approval. It also works before meetings, producing daily and meeting-specific briefs from your calendar and company context which is especially useful when you're stepping into someone else's meeting cold. As new emails, messages, documents, and meetings arrive, the shared knowledge stays current, and AI agents can reach it through an MCP server when drafting documents or doing other work.

The core problem is lack of context and memory with helping people and their agents know what's been discussed, what needs doing, and what the rest of the company already knows. The team is hunting for a more specific customer niche, but the underlying product works for essentially any computer-based team.

Under the hood
Boring on purpose. TypeScript, Next.js, and plain Postgres with pgvector. Recall.ai handles meeting capture, Google's gemini-embedding model powers search, cheaper Gemini Flash models do the bulk AI work, and Claude models are reserved for review and complex tasks.

The early signal came from close to home: multiple teams from ruum reached out, keen to start using Erra. Many arrived battle-scarred and there's real value in having someone walk you through 20 different knowledge-base products and explain precisely why none of them actually worked.

The Hard Part

Kristofer is refreshingly blunt: the hardest chapter is right now. erra is six weeks old, and they've spent the last three unable to develop the application at all because they have to rebuild their Google Cloud infrastructure and documentation, the direct cost of moving fast without enough due diligence early on. New to the scene and facing more possible niches than their experience can confidently rank, they've questioned a few things.

That connects to the assumption that broke first: that validating a niche would be easy. Everyone told them to pick one and build from there; nobody warned them how hard picking would be. They've now seriously considered more than ten different markets. The lesson is being learned in real time and their stance toward it is the tell. "We're not afraid to fail," Kristofer says, "and because of that, we're not afraid to push the limits." They're treating the whole stretch as a learning opportunity and pushing through.

What's Next

Three years out, success looks concrete: becoming the best-known and most-used company for "brain" solutions in the Baltics, and earning the trust of governments and much larger corporations. Kristofer wants to bring a few more genuinely smart people onto the team while keeping it deliberately tiny.

The deeper ambition is structural. Building systems that can run most of the business itself, from product development to marketing to customer outreach, automating as much as possible and staying ruthlessly efficient. A small team running a big operation on top of the very context engine they're selling.

The Toolbox

The obvious daily drivers are Claude Code and Codex. Kristofer estimates 90% of the work happens there, from outreach and social to building the product and brainstorming. The more interesting pick is Willow Voice (which he used to answer these very questions): prompting AI by voice, he's found, is far faster than typing and gives the model better context, with a generous free tier and noticeably more speed than other voice tools he's tried. His non-technical tool is the oldest one there is. Talking through the work with a partner or friend who has no background knowledge, purely to stay on track.

On recommendations, he's honest to a fault: most of what he's absorbed lately has come through carefully curated Instagram Reels, his feed tuned to entrepreneurship, web dev, and startups as his attention span for books has faded. The exception is recent and pointed. He's been reading about marketing because distribution is becoming one of the biggest moats for new startups. Purple Cow was short, but it got him thinking creatively about how to stand out, and he's hoping to apply some of it at launch.

🧠 How you can help
erra's next step is onboarding a few small teams for pilots for collecting real feedback and iterating fast. If you run a small, computer-based team (bonus points if you work across Baltic languages or feel the "nobody remembers what we decided" pain acutely) and want to help shape an early product, the founders want to hear from you. Reach out via Linkedin.