Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

Minute / Local Transcribe

The governed thing and the measured thing: a state-built meeting scribe

A government-built scribe transcribes council meetings and drafts standardised summaries, and every council in the cohort runs the same shared instance under one pooled assurance record. Modeled on the Minute / Local Transcribe program. The assurance record is thorough about data protection and process and silent about accuracy. Watch the split, where the governed thing and the measured thing come apart, and the one-to-many loop where a single instance's defect could surface across every council at once.

Stylized model of a documented deploymentCaseworker documentation & copilots

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the Minute-class pooled-assurance meeting scribe network: 6 components and 14 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 4 assumed · 2 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

  • assumed

    This models the state-built, cohort-governed meeting-scribe pattern documented in the Minute / Local Transcribe case file - not a reconstruction of the actual tool.

  • baseline

    The defining dynamic is a split record: the pooled cohort assurance is documented and strong on process (a shared data-processing template reviewed by the department as data processor, shared impact-assessment templates, bi-weekly cohort calls most councils rated useful) while no transcription-accuracy or error-rate evaluation of the shared instance is published, and standard risk controls such as penetration testing and certification were not completed on the alpha at pilot time. Those are the latent accuracy-check pathways levers can open.

  • baseline

    The one-to-many coupling is the modeled shape: a single centrally hosted instance transcribes for every council, drawn here as one model writing into two case-record stores plus a model-to-model correlated-defect pathway, so a single tool-level defect surfaces across many records at once. This is a state-built, centrally hosted arrangement with no commercial vendor and no data-processing agreement to an outside processor, so there is no vendor-egress pathway - the opposite of the bought-copilot shape.

  • assumed

    Peer pathways are authored on both signs: prompt templates and summary practice spread across the shared cohort hub, one shared instance homogenizes the record's phrasing across councils, and the documented bi-weekly cohort calls are a real community check that surfaces issues; both accuracy checks (a standing accuracy read, and a cross-instance defect check) start closed - the only published scrutiny is sector-wide, not tool-specific.

  • assumed

    The independent sector research raises risks of bias and hallucination for people who draw on care. Those served people are not on this diagram and never in these dynamics; that concern, and any pattern in who it affects, are recorded in the case file and measured outside any diagram like this one.

  • assumed

    Every time-savings figure cited for this tool (halved note-taking, up-to-90 percent recap reduction, about one hour saved per one-hour meeting) is self-reported by users or asserted by government, not independently measured, and no transcription-accuracy evaluation is published; this Lab models institutional propagation only and estimates none of them.

What this example does not show

  • The independent sector research raises risks of bias and hallucination for people who draw on care. The Lab models institutional propagation only, no demographics and no differential harm to served people; that concern is recorded in the case file and measured outside any diagram like this one.
  • Every time-savings figure cited for this tool is self-reported by users or asserted by government, not independently measured, and no transcription-accuracy evaluation is published. This shape models how the tool moves through council workflow and estimates none of those quantities.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • The UK government built its own AI meeting scribe for council caseworkers and piloted it through a cohort of 25 selected councils (22 active, more than 400 users) under one shared pooled-assurance record, then open-sourced it and adapted it to enlist around 500 housing and homelessness workers by June 2026; the cohort published a multi-council governance dataset but no transcription-accuracy or error-rate evaluation, and standard risk controls such as penetration testing and certification had not been completed on the alpha at pilot time.

    empirical
    • Government Local Government Association, Community led innovation in local government: Insights from the Minute pilot (2025) https://www.local.gov.uk/publications/community-led-innovation-local-government-insights-minute-pilot
    • Government Local Government Association, Artificial Intelligence Update (People and Places Board, 11 June 2025) https://lga.moderngov.co.uk/documents/s50505/Artificial%20Intelligence%20Update.pdf
    • Government Ministry of Housing, Communities and Local Government, Introducing Local AI (MHCLG Digital blog, 2026) https://mhclgdigital.blog.gov.uk/2026/03/16/introducing-local-ai/
    • Trade press Trendall, MHCLG enlists 500 council workers to progress work on AI transcription tool (PublicTechnology, 2026) https://www.publictechnology.net/2026/06/11/communities-housing-and-planning/mhclg-recruits-500-council-workers-to-progress-work-on-ai-transcription-tool/
    • Government Incubator for Artificial Intelligence (i.AI, UK Government), Frontline Services, Caddy (programme page, 2026) https://ai.gov.uk/our-work/frontline-services/

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Deadline pressure — Statutory or managerial timeliness rules reward fast approval of machine output over slow disagreement.
  • Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.
  • Staff turnover — Experienced skepticism leaves; new staff calibrate their trust on the tool itself.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Caseworker documentation & copilots domain page.

Levers available here and the patterns behind them

Documented case histories