Project approach

Problem selection: Phase One criteria

We have identified a long list of possible projects across several sectors. A project is selected for Phase One only if it meets all three of these tests:

  1. A clear boundary. The scope of the problem can be defined and agreed, and it can be solved without other problems having to be solved first. If the person responsible for the problem cannot describe its scope in one conversation, it is not ready for us to take on.
  2. Can be finished. A working prototype can be delivered within Phase One with the tools, skills and budget available to us.
  3. Can be owned by the client. The client can own the solution outright, and it can be documented well enough that the client’s own staff, with training and, where needed, recruitment support, can take on more of its day-to-day running over time. We continue to support it for as long as the client needs us.

Compliance and safety by domain

Each sector has its own legal and regulatory requirements. We confirm which apply before a project starts, and the cost of meeting them is included in the project budget from the outset.

Health and social care

Any solution that processes patient data, influences clinical decisions, or operates within a regulated health environment must comply with applicable clinical technology assessment standards. If the solution could directly or indirectly affect patient care, it requires a clinical risk management assessment under applicable clinical safety standards, and a qualified clinical safety officer must be engaged for the project.

This compliance work is included in the timescale and budget of each health project. A tool that sorts, prioritises or classifies patient-facing information may be regulated as a medical device, even if the underlying technology is simple.

Data processing in health projects must comply with patient confidentiality principles and health-sector information governance frameworks. Registration for those frameworks is part of Phase One setup.

Emergency services

Projects involving emergency service data, such as call recordings, dispatch information and incident reports, involve highly sensitive personal data and decisions made under time pressure. Systems are designed so that a fault does not put anyone at risk, and the risk assessment covers what happens if the system gives a wrong result, responds slowly or stops working.

Local authority and civil service

Projects handling personal data for local authorities must comply with applicable data protection law and any sector-specific codes of practice. Data Protection Impact Assessments (DPIAs) are a standard deliverable. Solutions involving decisions about individuals, such as benefits, housing or social care, need particular care over fairness in automated decisions, being able to explain how a decision was reached, and the right to have a person review it.

Defence and security

Projects involving classified data or defence applications require security clearance for all personnel involved. In Phase One we take on unclassified defence work only. In Phase Two we intend to apply for the accreditation an organisation needs to work at higher classifications.

Responsible AI use

Some projects involve artificial intelligence (AI) or machine learning components. These are the rules we work to:

Illustrative example: small build (software only)

Local authority document router

This and the following examples are illustrations of how we would approach a project, not descriptions of delivered work. A district council receives several hundred pieces of correspondence a week (letters, emails, web form submissions) and routes them by hand to the correct department. Routing is slow and inconsistent, and depends on knowledge held by two members of staff who are approaching retirement.

We would scope a classification tool that routes incoming correspondence to the right department queue with a confidence score, and flags anything below the threshold for human review. It would use a language model adapted for this task and hosted on the council’s own servers. The build would take a few engineering days and be tested against six months of past correspondence, with scoping, review sessions and handover spread across a couple of months. The council would receive the software itself, documentation and written instructions for its IT team, and we would continue to support it after handover.

Cost to client: approximately £6,000. Estimated annual saving: £40,000, from staff time released for other work and shorter processing delays. No personal data would leave the council’s systems, and there would be no ongoing licence fees.

Illustrative example: medium build (health adjacent)

Social care pathway case management

A local authority needs a digital case management system for a specific social care pathway. The current system is a spreadsheet maintained by one person who is due to retire. The data includes personal information about vulnerable adults.

Over about three months we would scope, prototype, test and deliver a web-based system with data governance, a record of who has viewed or changed each case, access controlled by staff role, and user training, plus a DPIA. We would host it on the authority’s own servers, on ours or with a hosting partner, whichever suits the authority. After handover to its IT team our support would continue, and we would help recruit one junior developer to take on day-to-day maintenance in-house.

Cost to client: approximately £35,000. The new system would record who has viewed or changed each case and limit access by staff role. A spreadsheet of personal data about vulnerable people without these controls is a data protection risk, and a breach could lead to a six-figure fine from the Information Commissioner’s Office.

Illustrative example: medium build (emergency services)

Call transcript analysis for mental health indicators

An emergency services body needs to identify which incoming calls have a probable mental health component requiring specialist intervention. This is currently done by hand and inconsistently, and as a result some responses are delayed and some calls are not escalated when they should be.

This would be scoped as a three-month medium build. The timescale covers clinical safety assessment under the applicable standards, with a clinical safety officer engaged as an associate, failure mode analysis, a data protection impact assessment (call recordings are highly sensitive personal data) and testing that it works correctly with the body’s existing dispatch systems.

The tool would transcribe and classify incoming calls in near real time, flagging probable mental health indicators for triage review. It would be advisory only: no call is routed or deprioritised without human review. It would be hosted on the body’s own servers. The handover would cover the software itself, documentation, the clinical safety case and the data protection impact assessment, with two of the body’s staff trained to look after it. Our support would continue after handover.

Cost to client: approximately £45,000 including clinical safety officer engagement. Expected benefit: faster specialist response and fewer calls missed for escalation. Any savings figure would depend on a baseline measurement agreed during scoping.

Illustrative example: small build (hardware and software)

Environmental sensor network prototype

An environmental monitoring programme needs low-cost air quality and water quality sensors deployed across remote sites with no mains power or reliable connectivity. Commercial sensors cost thousands of pounds per unit and only work with the manufacturer’s own online data service.

We would prototype a sensor unit built from low-cost, widely available components, solar powered and using long-range, low-power radio to send its readings, with openly licensed (open-source) software to collect and store the data. Ten prototype units would be built, tested in the field and handed over with circuit diagrams, the software that runs on each unit, and documentation for the data collection system. We would arrange hosting for that system, and train the client’s technical staff to build further units and to look after it.

Cost to client: approximately £8,000 for ten prototype units plus the data collection system. Commercial equivalent: £50,000 or more, plus ongoing platform fees.