Could legal costs become harder to recover when AI should have been used?

For the past few years, almost every professional discussion in the medico-legal space about artificial intelligence has focused on the risks of using it. What happens when AI produces an inaccurate answer? Who is responsible if confidential information is mishandled? Can a lawyer, doctor or expert safely rely on an output they do not fully understand the working behind? These are important questions, and they are not going away. But the UK Jurisdiction Taskforce’s Legal Statement on Liability for AI Harms, published on 7 July 2026, raises a different question, one that may ultimately prove just as significant: Could a professional be criticised for failing to use AI when a reasonably competent professional would have used it?

The debate is beginning to shift

The Taskforce’s legal statement considers how existing English law might apply when AI causes harm. One of its more striking conclusions is that professional liability may not be limited to situations where AI has been used badly. There may also be circumstances in which failure to use an appropriate technological tool falls below the standard expected of a reasonably competent professional.

That does not mean lawyers are now required to use AI, and it certainly does not mean that every task should be automated. The test remains grounded in ordinary professional standards. Was the technology reliable? Was it suitable for the task? Was it reasonably affordable and readily available? Would a responsible professional in the same position have considered using it?

We expect the answer will vary between cases and will change over time. A tool that is experimental today may become routine tomorrow. A process that is reasonable when no dependable alternative exists may become difficult to justify once safer and more efficient methods are widely available. That is the point we believe deserves closer attention.

What might this mean for legal costs?

The guidance does not say that legal costs are automatically irrecoverable when AI has not been used. There is no new rule requiring solicitors or experts to prove that they used artificial intelligence before their fees can be recovered. Nor are we aware, to date, of any reported English medico-legal costs decision in which charges have been disallowed specifically because an AI system was not used. It would therefore be wrong to suggest that the legal position has already changed that far.

However, courts already ask whether costs were reasonably and proportionately incurred. They can reduce costs where too much time was spent, work was duplicated or tasks were performed at an unnecessarily senior level. As specialist technology becomes more capable, it is easy to see how the way a task was carried out could become part of that assessment.

Imagine a clinical negligence case involving 8,000 pages of medical records. A paying party might reasonably ask:

• Why were dozens of hours spent manually sorting and searching those records?
• Why was routine document review undertaken by a senior fee earner?
• Why did the solicitor, paralegal and medical expert each repeat similar parts of the review?
• Could an appropriate system have organised the material, identified key entries and produced a preliminary chronology more efficiently?

These questions do not assume that AI should replace the lawyer or expert. They ask whether professional time was used appropriately and proportionately. The issue is unlikely to be whether the work needed to be done. The issue will be whether the process used to complete it was reasonable and proportionate.

Why medico-legal cases are particularly exposed

For those working in clinical negligence, personal injury and medico-legal practice, this is not an abstract question. It goes directly to the way cases are prepared, how professional time is used and, potentially, whether all that time should be recoverable from the opposing party. The risk is clear to see in medico-legal cases involving large medical-record bundles. Clinical negligence and serious personal injury claims often require thousands of pages to be:

• organised and paginated;
• separated by provider and document type;
• reviewed for duplicate and missing records;
• searched for relevant consultations and investigations;
• arranged into a chronology;
• revisited repeatedly as the case develops; and
• prepared for consideration by one or more medical experts.

All this work is necessary. But it is also repetitive, time-consuming and invariably expensive. Historically, there has often been no practical alternative to manual review. A human being had to work through the records page by page, but that position is changing.

Specialist systems can now assist with document classification, search, chronology generation and evidence extraction. They can help professionals locate relevant information and return to the original source material quickly. This does not remove the need for professional review. It may, however, significantly reduce the number of hours spent on work that is primarily administrative or information-processing in nature. As these tools improve and become more widely adopted, a traditional manual process may no longer be accepted as proportionate simply because it is familiar.

This is not an argument for replacing professional judgment

At ClinLexis we have been consistent about our position. AI should not decide whether there has been a breach of duty. It should not determine causation, assess prognosis or replace the independent opinion of a medical expert. Those are matters of professional judgment.

There is a fundamental difference between locating and structuring evidence and interpreting what that evidence means. AI may help a lawyer or expert identify a consultation, investigation result or medication change. It cannot assume responsibility for deciding whether that event is clinically or legally significant. The professional must remain in control.

That means checking outputs against the original records, understanding and accepting some limitations of the system and correcting any errors. It also means being able to explain how the technology was used and what safeguards were applied. The appropriate model is not therefore unsupervised automation. It is technology used to support a professional who remains responsible for the final work.

Not every AI system is suitable

It would also be wrong to suggest that the mere existence of an AI tool means it should be used. A firm should not be expected to upload confidential medical records into a public consumer chatbot simply because it is fast or inexpensive. Before relying on a system, professionals should ask:

• Is it secure?
• Is it suitable for confidential medical information?
• Can its outputs be traced back to the original records?
• Does it provide an audit trail?
• How accurate is it in practice?
• Can errors be identified and corrected?
• What happens to the data?
• Who remains responsible for the final work?

These are not secondary considerations. They are central to whether the use of the technology is professionally defensible. There will also be cases where manual review remains appropriate. A short bundle may not justify the use of specialist software. A particularly complex evidential issue may require close reading from the outset. A system may not be sufficiently reliable for a particular document type or clinical specialty. The decision should depend on the circumstances of the case, not on enthusiasm for technology.

A more realistic future question

For some tasks, the answer will still be a traditional manual process. For others, firms may increasingly need to explain why a large amount of professional time was spent on work that could have been supported by secure and well-established technology.

This is when the recent guidance really starts to matter. It suggests that professional standards may evolve in both directions. Using AI carelessly may expose a professional to criticism. But refusing to consider appropriate technology may eventually create risks of its own.

What should firms be doing now?

We do not believe firms need to rush into adopting every new AI product. They should, however, begin making conscious decisions about where technology might improve the way cases are prepared.

For medico-legal firms that means identifying the most repetitive and document-heavy parts of the workflow. It means testing specialist systems in a controlled way, measuring whether they genuinely save time and ensuring that human review is built into the process.

It also means being honest about charging. If technology significantly reduces the time needed to complete a task, it would be difficult to justify billing as though the work had been performed entirely manually. At the same time, responsible AI use has its own costs. Systems must be purchased, tested, governed and supervised. Outputs must still be checked; a human must remain in the loop because professionals will remain accountable for the final work product.

The objective should not be to remove professional involvement or simply drive fees down. It should be to ensure that highly trained professionals spend more of their time applying judgment and less of it performing repetitive document handling.

Our view at ClinLexis

We believe the discussion about AI in legal practice has sometimes become too polarised. On one side, AI is presented as a threat that cannot be trusted. On the other, it is described as a solution that will replace large parts of professional work. Neither position reflects how we believe this technology should be used.

The better approach is more measured. Use AI for the tasks it can perform well. Do not use it for judgments it cannot safely make. Keep professionals in control. When applying AI to medical evidence specifically, we need to preserve access to the source evidence and ensure that processes remain transparent and auditable. Importantly, we must be able to explain why the chosen approach was appropriate for the case.

There is not yet a general legal duty to use AI in medico-legal work. Nor is there a rule making manually incurred costs automatically irrecoverable. But the direction of travel is becoming clearer. As safe and reliable systems become part of normal professional practice, firms may increasingly be asked not only whether their work was necessary, but whether it was carried out in a reasonably efficient way.

At some point, “this is how we have always done it” may no longer be a sufficient answer.

This article represents the author’s perspective and is provided for general information only. It does not constitute legal advice.