[co-author: Alex Cumming]
When Sir Brian Leveson published Part I of his Independent Review of the Criminal Courts, the emphasis was on structural reform. Part II moves decisively into operational reforms, setting out a dense and practical package aimed at getting the system moving at speed. Across the review, one theme stands out more strongly than in Part I: the expectation that technology, including artificial intelligence, could become part of the everyday machinery of criminal justice. AI appears repeatedly as a capacity tool, supporting charging decisions, listing frameworks, disclosure schedules and transcription.
Leveson also reiterates a key point from Part I. His recommendations were never intended to operate as a “pick-n-mix”, and selecting a small subset of interventions would, in his view, be insufficient. Meaningful improvement requires a whole system approach in which structural reform, efficiency measures and technology adoption operate together. Part II is an attempt to provide the operational detail needed to underpin that integrated approach.
For corporates and regulated sectors, the relevance of Part II is practical. The review targets the parts of the system that typically increase costs and delays for organisations involved in criminal or quasi-criminal processes. These include early investigative decisions, sprawling digital disclosure, inconsistent listing practices and delays caused by avoidable logistics. The proposals, if adopted, could reshape both the pace and the nature of contested proceedings.
Strengthening investigations and charging decisions
A major focus of Part II is on improving the quality and timeliness of early investigative decisions. Leveson identifies poor case file quality, inconsistent communication and a less experienced workforce as key contributors to delay. For corporates, these weaknesses can translate into prolonged uncertainty, extensive legal costs, witness churn and increased long-term data retention burdens.
Part II proposes a range of measures to tighten early-stage decision making, including stronger collaboration between the police and the CPS, improved joint training and clearer guidance.
Leveson also recommends deploying AI-enabled support tools for charging decisions. These tools would not replace human judgement but could reduce administrative burdens.
Disclosure: reducing volume and improving discipline
Disclosure remains one of the most complex areas of criminal litigation. The growing volume of digital evidence has made existing processes inefficient and, in some cases, risky. Part II proposes removing the rebuttable presumption that certain categories of material automatically satisfy the disclosure test (for example, material that may cast doubt on the accuracy of prosecution evidence). The intention is to change disclosure so that the focus returns to relevance rather than volume. However, this shift may attract resistance from defence practitioners, who may view the removal of the presumption as raising the threshold for what must be disclosed and, in practice, limiting access to potentially important material.
The review also supports the use of AI summarisation tools for disclosure schedules. These tools could help organise large datasets, enable courts to understand scope more readily and reduce administrative friction. Taken together, these proposals reflect a move towards a more formalised systems and controls model.
Listing and case progression
Part II proposes several adjustments aimed at reducing the number of ineffective trials. In the Crown Court, Leveson recommends a national listing framework supported by AI-assisted scheduling tools and real-time dashboards within the Common Platform. This could improve visibility of capacity and allow cases to move between courts more readily.
The review also emphasises stronger case progression. Recommendations include appointing Case Progression Officers in all courts, simplifying and digitising the Criminal Procedure Rules so they can be integrated into digital tools, and expanding automation for administrative tasks. National adoption of Final Review Hearings is also proposed to help ensure that cases listed for trial are genuinely trial-ready.
For organisations and senior individuals involved in proceedings, the intended benefits include fewer wasted hearings and clearer timelines.
Technology and AI: becoming part of standard process
Technology appears throughout Part II as a core element of Leveson’s proposed throughput strategy. He recommends expanding remote participation for preliminary hearings in the Crown Court and for certain professional witnesses. Trials should remain in person, but remote attendance for police and continuity witnesses would be the default.
The review also proposes piloting AI-enabled simultaneous interpretation to address interpreter shortages, with clear boundaries around its use. AI-assisted transcription in magistrates’ courts, enhanced digital tools for case progression and improvements to data interoperability round out the package.
Leveson has since stressed the scale of the digital evidence challenge, warning that law enforcement agencies and prosecutors face “disaster” if they fail to use AI to process and review material. Without such tools, he cautions, many large and complex cases may simply never reach trial.
Conclusion
Part II is designed to turn the structural ambitions of Part I into an implementable operational plan. However, Leveson is clear that isolated interventions will not be enough, and that the system needs coordinated structural reform, technology adoption and increased efficiency if it is to regain momentum. For corporates, the implications are significant. Expectations around digital capability, procedural discipline and engagement with AI supported processes would rise sharply if the proposals are implemented.
The government has showed support for the broad thrust of Leveson’s recommendations and introduced the Courts and Tribunal Bill on 25 February, reflecting the reforms announced by Lammy last year. As with Part I, however, the real test will be delivery. Whether Whitehall can translate this operational blueprint into meaningful improvements in speed, certainty and system performance remains an open question.
[View source.]
