England and Wales have one of the worst court backlogs in the developed world right now. By the end of 2025, around 80,200 cases were sitting unresolved in the Crown Court, the part of the system that handles serious criminal cases. That figure is more than double where it stood before the pandemic. Government modelling from mid-2025 projected the number could reach 104,000 by 2029 if nothing structural changes.
The human cost of that is real. Victims of serious crimes, including rape and sexual assault, wait years for trials. Defendants sit on remand. Witnesses' memories fade. Cases collapse. The backlog of sexual offence cases alone grew by 40% between 2023 and 2025.
The government's response includes several measures, with an AI scheduling tool called J-AI at the centre of the technology piece. The tool is designed to help judges make better decisions about how cases are listed: which ones to push forward, how long each one is likely to take, and where gaps in the court calendar can be filled. The technology is adapted from a system already used in the NHS. If the pilot goes well, the government plans a national rollout to replace the current process, which largely runs on basic spreadsheets done separately at each court.
This is a genuine operational problem that AI is well-suited to help with. Scheduling is exactly the kind of repetitive, data-heavy task where pattern recognition tools can add real value. The most common reason for trials failing in Crown Courts in 2025 was overlisting: too many cases booked into slots that could not actually accommodate them. A smarter system that estimates hearing times more accurately could, in theory, reduce that meaningfully.
But the context matters, and the legal profession is making sure it is not lost. The Law Society put it plainly: AI can ease some administrative pressure, but it is not a replacement for investment in court buildings and additional staff. That is not a conservative reaction to new technology. It is an accurate description of why the backlog exists in the first place.
The roots go back to 2008. After the financial crisis, the Ministry of Justice saw its budget fall by around 12% in real terms over the following decade. Dozens of courts were closed. Legal aid rates for criminal lawyers were effectively frozen for years, making the work financially unsustainable for junior barristers. England and Wales have roughly three professional judges per 100,000 people, compared to a European average of around 22. In 2022, criminal barristers went on strike. There is also a maintenance backlog in the court estate itself worth around £1.3 billion.
The government is spending more. Courts will receive their highest-ever level of funding in 2026/27, and the cap on Crown Court sitting days has been lifted. A further £287 million is going into fixing court buildings. These are real commitments. But the AI scheduling tool is not replacing any of that work: it is being layered on top of a system that needs all of it.
There is also a legitimate caution around AI in legal settings specifically. Globally, courts have logged over 1,200 documented cases where AI tools produced false information that was submitted as fact. In the UK, police forces have faced scrutiny after an AI tool generated a reference to a football match that never took place, which then influenced a real operational decision. The Law Society has flagged that AI handling of witness statements, for example, risks missing cultural or linguistic nuances in ways that could affect outcomes.
J-AI as currently designed is doing something much more contained: helping manage a diary, not interpreting evidence or advising on guilt. That is an important distinction. Used carefully, within those limits, it is a sensible tool. The risk is not the tool itself. The risk is that governments reach for AI as a visible, modern-looking response to a problem that also requires slower, less photogenic work: hiring, funding, building, and training.