Evaluation of AI-Assisted Evidence Presentation System (AiCES)
Summary of Brunel University of London Pilot Study report – August 2026
Brunel University of London conducted an independent pilot study comparing the AI Court Evidence System (AiCES) with conventional evidence presentation during simulated criminal jury trials. Four 90-minute mock trials were conducted using murder and theft scenarios, with each case run once using AiCES and once using conventional manual presentation. Barristers and judges participated alongside witnesses and jurors, creating a realistic courtroom environment.
AiCES uses its voice commands to retrieve and present documentary, audio and video evidence, thereby removing the need for advocates to manually operate presentation technology or rely upon a dedicated operator.
The study found that, when operating correctly, AiCES was capable of retrieving evidence extremely rapidly, frequently within only a few seconds. In the murder trial, the median retrieval time using AiCES was 10 seconds compared with 22 seconds manually. In the theft trial, AiCES achieved the fastest individual retrieval recorded during the study, at three seconds.
The pilot identified some reliability issues. A relatively small number of failures, including incorrect retrievals, failure to recognise commands, system restarts and repeated prompts initially created significant delays. Consequently, conventional presentation performed better overall in the theft trial and proved more consistent.
Brunel concluded that the principal issue was not the underlying speed or capability of AiCES, but the reliability and consistency with which that performance could be delivered throughout a trial.
The study also identified advantages beyond simple retrieval speed. Voice control reduced the need for advocates to divide their attention between examining witnesses and operating courtroom technology. AiCES was particularly effective with multimedia evidence, allowing CCTV and other recordings to be retrieved, replayed and their playback speed adjusted by voice command.
Brunel considered that these capabilities have the potential to improve the presentation and examination of complex digital evidence.
The researchers concluded that the problems encountered were largely technical implementation issues rather than fundamental limitations of AI-assisted evidence presentation and were capable of being addressed through further development. Priorities include improving reliability and system recovery, better recognition of natural speech and different accents, reducing unsolicited responses and simplifying voice commands.
Overall, Brunel described the findings as “encouraging but preliminary”. If the identified reliability issues can be resolved, AiCES could contribute to more efficient evidence presentation, shorter hearings, better use of court resources, cost savings and ultimately reducing delays within the criminal courts. The report recommends further controlled testing followed, if successful, by a carefully managed pilot within an operational court.
Addressing the points identified
Following the pilot and taking into account the issues raised in the evaluation, BGP reviewed the findings and recommendations following the proof of concept mock trials.
Using feedback from the mock trial participants, we worked with developers to improve the reliability, efficiency and functionality of AiCES as part of the development of a Minimum Viable Product (MVP). This work incorporated fixes and improvements addressing all of the recommendations contained within the Brunel report.
On the 11th September 2026, we demonstrated the latest iteration of AiCES to representatives from Brunel University and the barristers who had participated in the mock trials.
All parties agreed that the new version represented a significant improvement in the efficiency and functionality of the system. It also incorporated additional features developed in response to feedback received during the mock trials.
We are continuing to develop and refine AiCES with the objective of producing a reliable and efficient MVP capable of being trialled within the judicial process.
Should you require further information place contact Wilf Pickles.