Decision by Default: Mapping the Council Algorithms Missing From Public Minutes

The image of a council decision is still, for many people, a committee room: officers, members, a vote and a minute. But a growing share of local administration is shaped before any of that happens, by software that sorts applications, flags risks, ranks housing cases or triages benefits enquiries. The question for scrutiny is not whether these tools exist. It is whether anyone outside the town hall can see them.

The Ferret sent freedom of information requests to 120 UK councils asking about automated decision-making systems, related policies and procurement records. The responses do not describe a single, coordinated rollout. They describe a quiet and uneven adoption, with authorities at very different stages of understanding what they have bought and what it does.

The procurement paper trail

Procurement records are often the first place automated decision tools become visible. Contracts for case-management systems, eligibility checks, customer relationship platforms and risk-scoring software can include modules that perform automated triage without the contract description making that function obvious. In several responses, councils identified suppliers for wider service platforms but could not confirm whether automated decision features were switched on.

That is a significant gap. A council may know it has purchased a system. It may not know, or may not be able to show, how that system is being used in practice. The distinction matters because accountability depends on more than a contract. It depends on knowing what data goes in, what rules or models are applied, what human review exists and what happens when someone challenges a decision.

Policies in name only

Of the councils that responded, some had detailed policies setting out when automated tools can be used and who is responsible for reviewing outcomes. Others had no specific policy at all, relying instead on general data protection and administrative law duties. A further group had policies that referred to automation in broad terms without naming systems or describing their use.

This is not necessarily evidence of bad faith. Local government often operates through layered contracts and shared services, and a single council may use dozens of systems that include some element of automation. But the absence of a clear policy makes public scrutiny difficult. It also makes internal oversight harder, because officers may not have a common reference point for what counts as automated decision-making or when a human review is required.

Triage by spreadsheet and score

The uses identified in responses were often mundane, which is precisely why they can escape attention. Benefits teams use rules-based systems to sort enquiries. School admissions teams use software to apply oversubscription criteria. Housing officers use scoring tools to prioritise applicants. Social care services use risk indicators to decide who needs a fuller assessment. None of these functions is inherently improper. Each can improve consistency and speed.

The concern is that automated triage can become decision by default. If a system ranks cases, officers may focus on the highest-ranked cases first. If a model flags risk, that flag may shape the level of intervention. The human decision remains, but it is framed by the software. The public accountability question is whether that framing is understood, documented and open to challenge.

Commercial confidentiality and the missing audit

Several councils declined to release details of supplier systems, citing commercial confidentiality. Others released contract values but redacted the technical descriptions. In some cases, the authority said it did not hold a central list of automated tools, meaning the information would have to be gathered from individual service teams.

Commercial sensitivity can be legitimate. Councils should not disclose security-sensitive details or undermine a supplier's position unfairly. But confidentiality should not become a default answer to questions about public decision-making. If a system shapes access to housing, benefits or care, the public interest in understanding its role is substantial. The paperwork should be proportionate to that interest.

What scrutiny would look like

A more accountable approach would not require councils to abandon useful software. It would require them to keep a register of automated decision tools, describe their purpose and limits, and publish the human review arrangements attached to them. It would require procurement documents to state clearly when a system performs automated triage or scoring. And it would require councils to explain how residents can ask for a decision to be reviewed by a person.

Some authorities are already moving in this direction. Their policies show that the issue can be handled without paralysis. But the picture across the UK remains fragmented. The risk is not a sudden takeover by machine decision-making. It is a slow accumulation of software-shaped choices that nobody elected, recorded in contracts rather than minutes, and explained only when someone asks the right question.