Who it is for
Built for organisations drowning in AI ideas. Not just government.
GOVBRM was written from the receiving end of AI requests in public and regulated organisations. The problem it answers is not particular to government. Any organisation large enough, regulated enough or complex enough to receive more AI ideas than it can shape faces the same six questions, wherever it sits: central and local government, the NHS and healthcare bodies, financial services, defence, utilities, higher education, and any large or regulated enterprise, worldwide.
Imagine an organisation holding five hundred AI ideas at once. Which fifty are worth investigating. Which ten are worth funding. Which three should ever run with real autonomy. Who owns the value each one is supposed to produce. What governance switches on as the stakes rise. And, months later, whether the value promised at the start actually arrived.
GOVBRM answers all six questions with one method, whichever sector the five hundred ideas came from.
The same pressure, everywhere
A named pressure in every large or regulated organisation.
Each sector carries its own governance, its own regulator and its own language for it. None of that changes what GOVBRM does: it shapes the request, ranks it against everything else, and routes it into the governance a sector already runs, in the right order.
| Sector | Where AI demand piles up | What already governs it |
|---|---|---|
| Central and local government | Service delivery, casework, policy support, back-office automation | Appraisal such as the Green Book, service standards, departmental risk and assurance boards |
| The NHS and healthcare bodies | Clinical decision support, patient-facing tools, back-office and administrative automation | Clinical safety governance, information governance, care quality regulators |
| Financial services | Underwriting, fraud detection, customer service, advice | Prudential and conduct regulators, model risk management, board risk committees |
| Defence | Logistics, intelligence support, back-office and administrative systems | Security clearance and assurance regimes, sector-specific procurement rules |
| Utilities | Network operations, customer service, asset management | Economic regulator requirements, safety cases, critical infrastructure rules |
| Higher education | Research support, student services, administration | Research ethics committees, data governance, funding body requirements |
| Large enterprises | Every function at once, arriving through every channel | Internal risk committees, existing technology governance, board accountability |
| Any regulated industry | Wherever a wrong decision carries legal, safety or reputational consequence | The sector's own licence, standard or regulator |
Regulator and standard names above are UK examples, for a reader who wants something checkable. Substitute your own sector's licence, regulator or equivalent instrument; the six gates and the logic of the lanes carry across worldwide, as they already do for public bodies in the crosswalk.
A worked example, not a measured outcome
What happens to the five hundred ideas.
The six gates already do this work. The numbers below are illustrative, to show what each gate is for; no organisation has yet run all five hundred through GOVBRM and reported the result.
- Request. Five hundred ideas arrive, from every direction and in every shape: a supplier demonstration, an executive who saw something work elsewhere, a team already using a public tool. One route in, and none is turned away unrecorded.
- Shape. Each is tested for a real need beneath the request, a value range in the partner's own unit, and the lowest level of autonomy that would solve it. Most of the five hundred are not requests for a model at all; perhaps fifty survive as worth investigating further.
- Rank. The fifty are scored against each other and against everything else already in flight, on criteria a sponsor could repeat, not on who shouted loudest.
- Commit. Perhaps ten receive a named business owner and a committed number, stated as a range with the conditions for the upper half. This is who owns the value: a role on the organisation's own chart, not GOVBRM and not a supplier.
- Build. Governance switches on in proportion to what could go wrong, not to how the request arrived. A handful of the ten are low risk and move fast; the rest carry more assurance, and perhaps three are ever pitched at real autonomy, an agent acting with limited human checking, because most tasks are solved two rungs below that.
- Review. Nine months after go-live by default, measured value is put beside the committed number for every item that reached Commit, whether it was one of the three or one of the ten. That is whether the promised value arrived, and it is asked every time, not only when the answer is likely to be good news.
Why this is not a hypothetical problem
Organisations can reach for AI more easily than they can decide what it is for.
In February 2026 the UK Department for Science, Innovation and Technology published research from 3,500 interviews with UK businesses, carried out between February and May 2025. Seventy-one per cent named not having identified a use for AI as a barrier, the most commonly cited answer of all. Sixty per cent cited limited AI skills or expertise. Among businesses that had already adopted AI, only twelve per cent reported an increase in revenue.
Read together, the figures describe organisations that can reach for AI more easily than they can decide what it is for, staff it, govern it or check afterwards whether it paid off. That is a shortage of practical mechanisms for prioritising, governing and measuring AI demand, not a shortage of access to AI tools. It is the gap GOVBRM is built for.
The figures are the department's own, describing UK businesses generally. They are not a claim about GOVBRM's own results: no organisation has yet run GOVBRM and reported its outcomes. See the provenance page for what is established, what is original to GOVBRM and what is still a hypothesis.
About the name
It reads as public sector, because that is where it began.
AI demand hit public and regulated organisations early and hard, with less room to experiment and more governance to satisfy, so that is where the method was worked out first, one request at a time. The name still carries that origin. What it names does not stop at government's edge: a regional bank, a hospital trust, a utility and a university all take in AI ideas faster than they can shape them, and the six gates ask the same questions of every one of them. What GOVBRM is and is not says plainly what it never replaces, in any sector.
Where to go next