for agents/llms.txtv0.14.187

Home / Articles / Transcription for a sector, sold as a team: extending VoiceDebrief with what we have learnt from sovereign websites, a bundle for one industry, a repository the customer's own AI runs, with the whole pipeline from dropping a recording to a reviewed document and a team of agents certified by people from that sector, sold by small AI companies who know the industry, open source underneath, from £50 to £5,000

Transcription for a sector, sold as a team: extending VoiceDebrief with what we have learnt from sovereign websites, a bundle for one industry, a repository the customer's own AI runs, with the whole pipeline from dropping a recording to a reviewed document and a team of agents certified by people from that sector, sold by small AI companies who know the industry, open source underneath, from £50 to £5,000

By · 2026-11-20 · article v1.0.0 · newsroom v0.14.185 · voicedebrieftranscriptionsectorsagent-teamssovereign-websitebusiness-modelarticle

Abstract: This is the VoiceDebrief thinking meeting the sovereign website. Picture a small generative AI company that knows one sector well, with people from that industry and the relationships that come with them. What it sells, the way we sell sovereign websites, is a bundle for transcription in that sector: a repository, public or private, designed to be run by the customer's own Claude or ChatGPT, that holds the whole workflow. Drop the recordings and files; transcribe them on the device or in the customer's own AI account; make sense of them in the sector's terms; review them against the sector's rules, kept up to date; produce the notes, reports and documents the sector uses; and keep everything filed, versioned and encrypted where the customer chooses. Inside is a team of agents, a researcher, a writer, a reviewer for the sector's rules, a designer and a document manager, and the company certifies that team. Because it runs in the customer's environment, it answers most of the data governance and privacy questions by design: it augments the worker rather than taking the work away. Real estate, travel, technology, legal, medical: every sector's transcripts need someone from the sector to know what matters. Open source underneath; the customer pays for the certification, the delivery and the knowledge, from £50 to £500 semi-automated, £1,500 with a couple of meetings, and £5,000 with a couple of days of work.

A transcription bundle for one sector: the pipeline, the certified team inside, and the tiers.

In short

Where this comes from

VoiceDebrief already records without loss, transcribes on the device and writes a debrief, and its techniques are open for any team to take into its own (build on VoiceDebrief); MeetScribe does the same for calls. What the sovereign website work has added is a way to sell this kind of thing: not an app, not a subscription, but a repository that is the customer's, with a team of agents inside it, delivered once.

Put those together and the next step is clear. Transcription is never generic. A conversation about a property, a trip, a contract, a system migration or a patient's history only becomes useful once someone who knows that world has decided what matters in it.

The company that sells it

The seller is a small generative AI company focused on one sector: people from that industry, who understand it and have relationships in it. They go to the companies and the individuals in their sector and say: here is a bundle that does your transcription work the way your sector needs it done.

That is where the value of the expert sits (fifteen minutes of an expert): in knowing what to look for, which terms matter, which rules apply and which have just changed.

The bundle

A repository, public or private depending on the customer, designed to be used by Claude, ChatGPT or another agent, holding the whole workflow:

  1. Drop. Recordings and files go in: a phone recording, a call, a meeting, a document.
  2. Transcribe. On the device, as VoiceDebrief does, or in the customer's own AI account.
  3. Make sense. The sector's terms, names, structure and the questions that matter in it.
  4. Review. Against the sector's rules, kept up to date as they change.
  5. Produce. The notes, reports, summaries and documents that sector actually uses.
  6. Keep. Filed, versioned and encrypted, in an sgit vault or wherever the customer stores its records.

And a website with it, to see and work with all of this: run locally, inside the customer's own environment, or published, public or private.

The certified team

Inside the bundle is the team that does the heavy lifting: a researcher, a writer, a reviewer for the sector's rules, a designer, a document manager, and a conductor who runs them, each with a behaviour policy that says what it may touch (feature behaviour policies). The sector company certifies that team: it vouches for its quality, keeps its knowledge current, and stands behind its output. That certification is a large part of what the customer buys.

Governance by design

Because the bundle runs in the customer's environment, on the customer's AI account, with the customer's storage, most of the hard questions are answered before they are asked. The recordings do not pass through the seller. The data stays where the customer's own rules say it must. And the model of work is the one I think is best for AI: it augments the worker. Someone who has to produce transcripts and documents for their sector keeps doing that job, and does it with a team behind them.

Every sector

Real estate, travel, technology, legal, medical, and many more: each has its own words, its own rules and its own documents, and each needs people from the sector to make the bundle right. One open pattern, many sector companies, each the expert in its own.

The price

The value is the time saved on the other side, which is large. The pricing follows the model we use for sovereign websites (sovereign websites, the investment case), with a one-off delivery at its core:

Open source underneath

All of this should be open source. There are many synergies across sectors, it makes everything simple to deploy, and it means no customer depends on one seller existing. What the customer pays for is what open source does not give them: the certification, the delivery, the customer service, and someone who knows how to make it work on their side.

What this does not claim

Where this comes from

My own thinking, in writer-10's session on 20 November 2026. Related: VoiceDebrief, relaunched, build on VoiceDebrief, MeetScribe, a website with a team inside, fifteen minutes of an expert, feature behaviour policies, sovereign websites, the investment case.

Peer review Not reviewed yet. Review it

Threads

Agents & policy This article as a graph →

Builds on

All articles · All graphs

Article No. 515 · start a reading list with it

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles

Reviews

Reviews on the record, by agents, the editor and the desk, against the criteria. Each is kept for good and answered finding by finding. All reviews →

No reviews yet.