Document Processing
Ofsted inspectors manually drafted children's home inspection reports from written inspection evidence, which consumed inspector time and created challenges around drafting speed, formatting consistency, and first-draft quality.
Reported impact
Ofsted expected the tool to produce structured first drafts faster, improve consistency and quality in draft reports, and free inspector capacity for evaluative work, but it retired the project after beta testing rather than progressing it further.
Ofsted built a report-drafting workflow to turn children's home inspection evidence into suggested wording for inspectors rather than a final regulatory decision. Inspectors enter their judgement in a web interface and submit written evidence gathered in OneNote and exported as PDF. The interface calls an API, and Python scripts act as the intermediary layer: they clean text by expanding common acronyms, sequence three report-drafting prompts, and run additional house-style prompts. The core generation step uses ChatGPT 4.1 Mini from OpenAI, hosted in internal Azure instances, with text held temporarily in RAM during processing. The system collates model outputs into a Word document covering three report sections, then passes that draft into Cygnum for the normal reporting workflow. Human inspectors remain responsible for review, editing, version control, and quality assurance, and the service does not retain input data for model training or long-term storage.
Inspectors collect evidence during children's home inspections, capture notes in OneNote, and export the evidence as a PDF. An internal web interface sends the inspector's judgement and PDF text through an API to Python scripts, which clean acronyms and orchestrate prompt calls to ChatGPT 4.1 Mini. The generated section drafts are collated into a Word document and uploaded into Cygnum for inspector review and QA.
"At present, the Python pipeline feeds a locally stored PDF document of text data into ChatGPT 4.1 Mini (an LLM by OpenAI) with some data cleaning to replace common acronyms with their full form."
"Following the beta phase of testing, Ofsted chose not to progress this project further - record updated to 'retired'."
"The project aims to drive efficiency in the inspection process by deploying a self-serve assistant for inspectors that creates suggested wording for 3 sections of a children's home full inspection report based on written inspection evidence using a Large Language Model (LLM)."
"The suggested wording for 3 sections of the report is output as a Word document that can be uploaded to our report management system Cygnum."
"These outputs are then collated in a Word document with a heading for each section. This document is then passed through a series of house styling prompts, applying rules to each section."
"Through a web interface these inspectors will provide a judgement alongside inspection evidence they gather in OneNote exported as a PDF. The web interface will connect to an API where Python will act as an intermediary for their text data to be processed by a Large Language Model (LLM) in OpenAI's ChatGPT family."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
OpenAI Sample Aug 31, 2026 Polimill Polimill needed to reduce administrative burden across fragmented Japanese municipal systems so public employees could find prior decisions, prepare assembly responses, and deliver services without each municipality… 3-5x development speed increase from Codex and OpenAI hands-on support Codex · GPT models · OpenAI API OpenAI Sample Apr 15, 2026 Commonwealth of Pennsylvania The Commonwealth wanted to improve the speed, clarity, accessibility, and accuracy of public services while reducing manual effort for state employees handling drafting, research, hiring support, document review,… 175 employees across 14 agencies Board of Pardons online tracking system · COMPASS · ChatGPT Enterprise OpenAI Sample Mar 21, 2025 Commonwealth of Pennsylvania Pennsylvania wanted to reduce time spent on drafting, research, summarization, IT support, procurement, contract analysis, and other administrative work across state agencies while keeping human oversight, transparency,… 175 employees from 14 agencies used ChatGPT Enterprise ChatGPT Enterprise 