Archives

Can AI Help Level the Green?

Public complaints, public money and a new player on the field

The guest post you are about to read from a good colleague takes a sideways look at suggestions that AI assisted complaints are creating impossible burdens on public bodies.

Mark Brooks

 

A recent BBC report described a rising tide of complaints and information requests written with the help of artificial intelligence. Councils, schools and other public bodies said that matters once set out on a single sheet of paper can now arrive as twenty pages of legal argument. Some contain irrelevant material, misquoted legislation or authorities that do not exist. All of that has to be read and checked.

Those are legitimate concerns. Anyone using AI remains responsible for what they send. A computer-generated thicket of words does not become a good complaint merely because it sounds legal. Yet the discussion also revealed something rather more important. For perhaps the first time, ordinary people have ready access to a small part of the procedural and drafting capacity that public bodies have always possessed. The public body calls the result an increased burden. The citizen may call it the first occasion on which anyone has taken the substance of the complaint seriously.

Professor Luke Clements describes the wider problem in Clustered Injustice and the Level Green. People who experience disability, poverty and other disadvantages are repeatedly required to play legal and administrative games in which the odds are stacked against them. The rules are complicated, the language is unfamiliar and the opposing player often helped to design the system. AI does not produce the level green that Professor Clements argues for, but it may remove a little of the slope.

The imbalance is particularly stark in disputes involving social welfare and public services. A council has professional officers, an internal legal department, access to external solicitors and counsel, years of records and policies, and the ability to conduct the dispute during paid working hours. The individual may be dealing with disability, caring responsibilities and financial pressure while trying to understand unfamiliar law at the kitchen table.

Legal aid was intended to prevent lack of money from deciding who could enforce a legal right. Its availability has narrowed severely, and specialist provision has become increasingly difficult to find. Community-care law has not vanished entirely from the legal-aid scheme, but for many people the possibility of obtaining timely specialist advice is more theoretical than real. Parliament’s Justice Committee concluded in 2021 that civil legal aid required a whole-system approach to reform. More recent research by the Institute for Fiscal Studies likewise records a substantial reduction in access to civil justice following the Legal Aid Sentencing and Punishment of Offenders Act 2012.

That absence of support changes the significance of AI. Used properly, it can explain a policy, locate the relevant statutory guidance, organise a long chronology, compare two versions of a document and help a person express the question that has been troubling them. It cannot replace legal advice. It can, however, help someone recognise when advice is needed and make far better use of it when it is available.

I have seen this in my own dealings with a local authority over adult social care, Direct Payments and disability-related expenditure. AI did not supply lived experience or decide what I believed. It helped me work through large quantities of correspondence, test assertions against the published rules and turn a general sense that something was wrong into precise questions. Crucially, it also helped with constructive work: drafting alternative policies and proposals rather than simply criticising the existing ones.

That distinction matters. A person who can say only that a decision feels unfair can be answered with a reassurance that procedure was followed. A person who can identify the relevant paragraph, explain the inconsistency and propose a workable alternative is harder to brush aside. The underlying grievance has not been generated by AI. AI has helped the person articulate it.

There is another public-money question that receives much less attention. Public funding for legal assistance to the individual is routinely treated as a scarce resource that must be tightly controlled. The money spent by a public authority resisting that individual is public money too, but it is rarely discussed in the same way.

Of course, councils must sometimes defend decisions and proceedings. A weak or opportunistic claim should not succeed merely because answering it costs money. But continued resistance is not automatically in the council’s interest simply because an officer has taken a position. The council’s interest is the lawful and economical administration of services for its community. That is not always identical to preserving every decision, defending every line of correspondence or avoiding an admission that something should have been handled differently.

This raises a legitimate governance question. Once a dispute becomes lengthy or expensive, who steps back and asks whether the expenditure still benefits the authority? Is anyone independent of the original decision reviewing the prospects, proportionality, cost to date and possibility of an early remedy? Or can public money continue to be spent because conceding the point would be uncomfortable for the department concerned? That is a question, not an accusation. It is also consistent with the statutory Best Value duty, which requires councils to secure continuous improvement and make effective use of taxpayers’ money.

Public authorities are themselves embracing AI for precisely those reasons. The Local Government Association reports that councils are adopting it to improve staff productivity, service efficiency and cost savings. Its AI practitioners’ network already involves hundreds of council officers. There is nothing improper in that. A council faced with lengthy correspondence could use AI to extract the issues, identify the evidence needed and detect recurring problems across complaints, provided that human judgment, confidentiality and accountability remain in place.

The difficulty arises if the technology is welcomed when it strengthens the institution but deplored when it strengthens the citizen. A council cannot reasonably celebrate AI-assisted efficiency in its own workforce while treating a resident’s responsible use of the same technology as inherently suspect.

There are practical ways to control the genuine problems. Complaints procedures can ask for a short initial statement covering what happened, why it is said to be wrong and what remedy is sought. Supporting evidence and detailed legal argument can follow when required. Complainants should verify every quotation and reference and remove material they do not understand. Public bodies should answer the substance rather than become preoccupied with whether a machine helped to express it. The standard should be accuracy, relevance and responsibility, not whether every sentence was typed without assistance.

AI will sometimes generate rubbish. So do people, lawyers and public bodies. The answer in each case is to test the material, not disqualify the speaker. Responsible AI can help councils resolve complaints earlier, identify poor practice and save money. It can also help a disabled person, carer or tenant understand a decision that materially affects their life. Those benefits are not in conflict.

For years, public bodies have possessed advantages of knowledge, language, time and professional support that most citizens could not hope to match. AI has not made the contest equal. It has made the imbalance more visible and given some people a better chance of being heard. If AI is beginning to level the administrative playing field, should public bodies complain about the new player, or examine why the field was so uneven in the first place?

Sources

.

Posted 25 September 2026