The facts

A PSV operator was subject to a maintenance-based investigation in relation to an S marked prohibition and a wheel-related incident caused by a loss of the nearside drive axle wheel pair.

Upon investigation the DVSA examiners alleged issues with maintenance including ineffective forward planners, driver defect reporting, exceeded inspection intervals, and lack of effective Wheel & tyre management systems in place.

Examples of the breaches found in the investigations include\;

• MOT fail rate of 21.84% compared to the national average of 11.40%
• Ineffective daily walkaround check of the wheel by the driver in relation to wheel loss incident

In response to the DVSA findings, the transport manager argued that they had previously had a satisfactory independent audit and has commissioned an expert report from a mechanical engineer to assist with understanding the wheel loss, who concluded that the least unlikely root cause was a catastrophic failed hub casting rather than simple wheel nut insecurity.

Due to these issues and ambiguities, the operator was called into a public inquiry before the Traffic Commissioner. JMW Solicitors were engaged to represent the Operator and Transport Manager at the Public Inquiry.

AI-assisted reports: the risks for expert evidence

The inquiry also highlighted the dangers of relying on AI-assisted drafting in a report presented as expert evidence. Concerns arose about the apparent use of AI in the wheel-loss investigation report and whether its conclusions could be supported by the underlying evidence. The issue was not simply how the report had been drafted (although the author has mistakenly left the usual AI prompts in the final version), but whether the opinions expressed could withstand scrutiny.

On the account of the hearing, the DVSA examiner was unable to substantiate material conclusions expressed in the report when challenged. That mattered beyond the wheel-loss analysis: it also undermined the weight that could safely be attached to the examiner’s evidence concerning the maintenance systems given that these were predicated on his view of the wheel loss incident itself. This does not mean that every recorded maintenance concern fell away, or that AI use alone rendered the evidence unreliable. Rather, it reinforced the need to distinguish objectively evidenced shortcomings from conclusions that lacked an adequate evidential foundation.

The practical lesson is clear: AI can assist with drafting and presentation, but a polished report is no substitute for a properly evidenced technical opinion. Anyone presenting expert evidence should check each substantive proposition against the source material, identify assumptions and limitations, and be able to explain the reasoning independently of the AI output. Operators should scrutinise the evidence supporting a report’s conclusions, rather than accept those conclusions simply because they appear in an official document.

The above applies equally to an operator who may use AI to assist in drafting responses to the DVSA or the Traffic Commissioner.

The Public Inquiry

The Traffic Commissioner concluded that it was more likely than not that the operator was not culpable for the wheel loss as there was sufficient evidence to support the notion that checks were conducted as required by the driver.

In relation to the unsatisfactory DVSA findings, the Traffic Commissioner found for the Operator, as general good compliance before the stated wheel-loss incident. The Traffic Commissioner therefore issued a warning to the Operator, with the repute of the Transport Manager retained but tarnished.

Undertakings have been given to the operators to carry out an audit of transport safety and compliance systems in April 2027.

Summary

The case highlights the importance of reviewing the evidence submitted by the DVSA and properly testing that evidence when required at Public Inquiry. It would have been easy to suggest that the Traffic Commissioner should simply disregard the AI generated report from the DVSA, but that wouldn't simply be enough, with the Operator still needing to present evidence to counter those conclusions.

The use of AI by all parties in proceedings is common, with it very much assisting those involved in drafting of documents. However, any report submitted to an inquiry, must be backed by cogent evidnece in support