# Transcript by Lightbloom AI: our AI for knowledge nobody has written down

Transcript is a tool we built for ourselves. When a problem depends on knowledge nobody has written down, it has a private conversation with everyone involved, asks follow-ups until the real process is on the table, and turns what people know into one picture of how the work actually happens. It is not a step every client goes through: we bring it in when the problem calls for it.

Live product: https://transcript.lightbloom.ai · Bring us a problem: https://cal.com/davidprittie/build · Contact: info@lightbloom.ai

## Everyone talks. One picture comes out.

Transcript hears everyone at the same time. It can ask one person about something others described, without ever naming them, keeps the contradictions, and asks the question nobody has asked yet. What people know but never wrote down reaches the model.

- Talks with everyone, in private. Everyone at once, each conversation as long as the person wants. Not a survey.
- Keeps asking. Follow-ups until the real process is on the table.
- Checks back with each person. Nothing counts until the person who said it has approved it.
- Builds one model of the company. A living picture of how it runs, kept current as the work changes. Your managers read it. We build from it.

## One conversation in front. A team of agents behind it.

Each person talks to one agent. Behind it, one decides what has been learned, one looks for contradictions across the company, and one tracks how much of each person's work is still unmapped and steers the next questions toward the gaps. That is why the questions get sharper as it goes, and why nobody is asked the same thing twice.

## What the picture holds

- Every process, as it really runs. How work actually gets done, not how the process document says it does.
- Every problem, priced in hours. What it costs the company in the hours your people spend on it. The largest return comes first.
- How your teams use AI today. Where it has taken hold, where it has not, and what that says about where to point it next.
- Every conversation, private. No one can read one, not a manager, not the management team. Everyone sees the model, never one person's words. Built into how the product stores data, not a policy.

## Works where the work is

Half of most companies never sits at a laptop. The same conversation runs on a phone, on the floor and in the field, so the whole company is heard.

## Questions people ask

**Who can see what I say?** Nobody. Your manager and the management team never see what you say. The product stores conversations so that no one can open them; that is built in, not a policy. Managers see how much of the company has been heard and what the company as a whole knows, never a conversation, a quote or a summary of one person.

**Does Transcript capture tribal knowledge before someone leaves?** Yes. That is most of what the model is made of: the way work actually gets done, which lives with the people doing it and rarely reaches a system.

**How is this different from process mining or an employee survey?** Process mining reads system logs, so it sees the steps that touched a system and misses the phone call, the spreadsheet and the workaround. A survey asks everyone the same questions once. Transcript has an ongoing conversation with each person about their own work, asks again where accounts disagree, and keeps the contradictions.

**Can it tell me who my single points of failure are?** Yes. The model shows who everything depends on: the handovers that run through one person, the report only one person can produce, the approval that waits for someone who is rarely in. You see the dependency, never the conversation.

**Is Transcript a hiring or assessment tool?** No. It describes how the work runs, not how well any individual performs.

**Does it connect to our systems?** It starts with your people, because that is where the real process lives, then your systems come in on top. Calendars first, and more as the model grows.

**Is my data used to train AI models?** No. Conversations stay within your organisation's own review, and your data never trains models.

**What does our team have to do?** A short note from the top saying this matters and that answers will not be used against anyone, and one named person who owns the rollout with us. Everyone is asked by default, on their own schedule.

**How long do you keep the conversations, and can we delete them?** Conversations stay in your organisation's own space while we work together. You can delete them, or ask us to, at any time. Built for the GDPR and the EU AI Act.

**Does every engagement start with Transcript?** No. Most start with a problem and a prototype in week one. We bring Transcript in when the problem depends on how the work really runs and nobody has written that down, and we say why.

## Then it goes into the build

Transcript is not sold on its own. What the company tells us changes what we ship that week, and Transcript keeps running while we build, so what surfaces later still shapes the work. https://lightbloom.ai/

More for agents: https://lightbloom.ai/llms.txt
