Mandatory reporting map
Owners, frequency, due dates and assurance status for NHS returns — with public reference links. Example: MHSDS monthly return row in the register.
View pageA practical demonstration of Business & Performance work
This is my personal demonstration site for the Business & Performance Business Partner role. It is not an official Dorset HealthCare website or report.
I built it to show how I would approach the work in practice — not just list skills on an application form. The examples walk through turning data into clear performance intelligence, checking assurance risks, supporting service improvement, and using AI safely under human control.
You can explore six draft briefs from public NHS data (for Dorset HealthCare as provider — RDY), a separate synthetic warehouse demonstration for fictional Demo Rivers Health (DRH), and pages on how bounded AI agents could support the work with governance and sign-off.
The role is not just about producing reports. It is about helping services understand what is happening, spot risks early, and make better decisions. A good Business & Performance Business Partner connects services, corporate teams, finance, information teams and senior leaders — and makes performance information useful, not just available.
The main idea
Agentic AI can help NHS business and performance teams work faster, more clearly and more consistently — when it is used with clear sources, human review and honest caveats.
In this demonstration, AI supports the work; it does not replace professional judgement. It can help with:
- drafting report text, as shown in the MHSDS access and activity profile
- checking figures against sources, as shown in the Report Analysis Agent conversation and the corrected urgent-care provider-month brief
- summarising trends, as shown in the Talking Therapies access brief (six-week access above standard but falling)
- explaining performance changes, as shown in the community services activity profile (month-on-month rise but six-month fall for key activity types)
- preparing meetings, as shown in the public statutory assurance source map (which sources exist and who to speak to)
- checking reporting rules, as shown in the mandatory reporting map and the public reports method
- organising evidence, as shown in the warehouse demo workflow and source profiling report
- reducing admin-heavy work, as shown in how agentic AI could help maintain a reporting register
- answering reporting technical and definition questions from approved sources, as shown in the agent operating model and MHSDS Expert Agent conversation
However, AI cannot and should not replace professional judgement.
People must still be responsible for definitions, data checks, interpretation, decisions, escalation and sign-off.
Purpose of this site
This site is a practical demonstration of how I would approach the role — shown through worked examples, not just described.
It shows how performance information can be made clearer (see the NHS Oversight Framework brief), how mandatory reporting can be organised and checked (see the mandatory reporting map), and how bounded AI agents could support busy teams when used safely (see the agent operating model).
The material falls into three kinds, kept clearly separate: draft reports built from public aggregate NHS data for RDY; a synthetic Demo Rivers Health (DRH) data warehouse example; and an agent operating model and governance explanation. RDY and DRH are different providers and must not be mixed up. Each page demonstrates a different part of the work:
Draft performance reports
Six agent-assisted briefs from public NHS data — e.g. the urgent care, diagnostics and beds source check explaining what each public source can and cannot support.
View pageAgent operating model
Bounded agents with approved sources and citations — e.g. the MHSDS trace example and Report Analysis QA example.
View pageGovernance and benefits
Benefits, controls and the AI assurance checklist — human sign-off before anything goes to services or Board.
View pageSynthetic data warehouse demonstration
Fictional DRH source extracts through profiling, warehouse design, SQL/pipeline specs and reporting QA — e.g. the warehouse human review pack showing where people must sign off before trusting synthetic outputs.
View demonstrationEach page starts with a plain-English summary, so the site can be followed without a technical background.
How this was built
This site was built using a Cursor agent, working from my prompts under human direction and review.
This is itself an example of responsible AI use — see governance and benefits for the controls that would apply in a Trust setting.
I set the structure, content and tone. The agent wrote the code, created the reports and supporting documents. I reviewed the outputs and directed changes throughout.
The site uses:
- static HTML, CSS and light JavaScript
- public NHS source references
- sample data for demonstration
- no backend database
- no connection to live Trust systems (all the work was done on my personal laptop)
What this site is designed to show
1. How I think about performance
Good performance reporting should explain what the figure is, what it should be, whether it is getting better or worse, and what action may be needed. It should help people understand the position, not just give them more numbers.
The NHS Oversight Framework brief shows trust-wide figure, peer position, trend and priority review flags — the kind of structured performance narrative a Board or oversight conversation needs.
2. How I think about assurance
Mandatory reporting needs clear definitions, source checks, trend analysis, known limitations and human sign-off. AI can help organise and test this work, but it must be clear where the information came from and what still needs checking.
The mandatory reporting map shows owners, frequency, due dates and assurance status (local fields are illustrative sample data). The statutory assurance source map shows which public sources contain RDY rows — it is a navigation aid, not a scorecard. The reporting-table assurance page applies the same discipline to a synthetic mart before anyone trusts it for narrative. The Warehouse report QA conversation shows the same flawed-draft → corrected-brief discipline on synthetic DRH data. The AI assurance checklist sets the human gate before sharing output.
3. How I think about service improvement
Good business partnering means understanding the real pressures services face — demand, capacity, staffing, pathways, finance, quality and patient care. Information should support improvement. It should not just create extra reporting work.
The synthetic urgent-care analysis triangulates contacts, cases, bank shifts and agency spend — and separates possible Jan–Feb operational pressure from a March extract-driven spike. After QA, the corrected provider-month brief (linked from the figure-checking example above) is what a service lead would see.
4. How I think about responsible AI
AI should be used carefully, with clear rules, human review, audit trails, data protection safeguards and honest caveats. The aim is not to use AI for its own sake. The aim is to give analysts, managers and services better support.
The agent operating model defines bounded, source-bound agents — not general chatbots and never the decision-maker. The MHSDS Expert Agent conversation traces a figure to source and refuses to over-interpret. The source profiling conversation shows bounded agents working on synthetic extracts only. Agent rules and governance and benefits document controls: no patient-identifiable data, human sign-off, version-controlled rules, and clear confidence levels.
Why this is relevant to the role
This site shows the mix of skills I would bring to the Business & Performance Business Partner role:
- NHS operational management experience
- performance and service improvement experience
- data, SQL and reporting knowledge
- understanding of NHS reporting and assurance work
- ability to explain complex information clearly
- practical experience directing AI agents
- commitment to transparency, accountability and public service values
Those skills are demonstrated in the worked examples above — for instance, performance narrative in the draft reports, assurance discipline in the mandatory reporting map and governance pages, and responsible AI in the agent operating model.
My background combines operational leadership and hands-on data work. That matters because performance information is only useful if it connects to the real pressures services face.
This is not a finished Trust product. It is a demonstration of approach: how I would use evidence, structure, governance and new technology to support better performance work.
Demonstration caveat: This is a personal demonstration site. It is not an official Dorset HealthCare website or report. It is not an operationally validated Trust product. It does not contain confidential, unpublished internal information or patient-identifiable information. Any figures, examples or draft reports shown here are for demonstration only. They use either public data or clearly marked sample data.
Explore this demonstration
If you have five minutes, try this route: browse draft reports, read the warehouse design conversation, open the MHSDS trace example, and use the assurance checklist.