A summary of the panel discussion hosted by CGI at this year's Modernising Criminal Justice 2026.
Speakers: Dave Watts is Director Consulting Expert in Justice at CGI. Lewis Lincoln-Gordon is Deputy Director for Enablement, Delivery, and Adoption at Police AI. Matt Bland is Chief Operating Officer of the Society of Evidence Based Policing and Craig Robb is Deputy Director for Crime at HMCTS.
Criminal justice has had reviews, recommendations, pilots, and reform programs for years. Dave Watts from CGI asked a simple question: why are known problems still so hard to fix? Why hasn't change happened? The answers, from experienced panellists across policing, courts, evidence and digital, were honest.
Craig Robb from HMCTS cited an academic theory that says when one bureaucracy tries to implement change, it has some chance of success. But when two or more bureaucracies collaborate, the chances of success diminish with each additional bureaucracy involved.
His working day at HMCTS tends to have tabs open with policing colleagues, CPS colleagues, probation colleagues, prisoner escort services, and judges. Add in policy across at least three government departments, plus various funding and accountability lines going upwards. "It's hardly surprising when you then add in policy across probably at least three government departments and all of the various funding accountability lines that go upwards that sometimes it takes a little bit of time for stuff to happen."
Historically, there was one division at the Ministry of Justice with people responsible for criminal justice strategy. Now there's a whole directorate with a Second Permanent Secretary to help join up the system. There's funding. There's direction. But this is brand new. The opportunity is only just arriving.
Lewis Lincoln-Gordon from Police AI described why policing has failed at previous reform attempts: a lack-of funding and buy-in from multiple governments. But there's now opportunity: the right language from government, funding starting to flow, and appetite for coordination.
Yet policing is by nature reactive. Responding to what's in front of your face. Being able to do strategic planning, forward thinking, collaborative work - that's where systems succeed. And that's hard. It's a leadership challenge, not a tech challenge. Leaders have to choose to think systematically rather than responding to immediate pressure.
Matt Bland from the Society of Evidence Based Policing spoke from outside the system, which gave him different perspective. The system is too fragmented to embed change sustainably. Too many departments with their own priorities, mandates, overlapping priorities that dilute efficiency.
Take digital forensics - 90% of criminal investigations now have some element of it. But the join-up between police forces is non-existent. There's a spaghetti diagram of national governance. And that completely ignores join-up to CPS, courts, the wider criminal justice system. The whole thing could be streamlined through one pipeline, but governance and accountability sit with separate organisations.
This is fundamentally an organisational leadership problem. And it's probably only going to get harder as organisations start leveraging AI and potentially solving problems in their own pockets. You need overall leadership across the system.
Craig pointed to something that's changed. The latest suite of reforms has a package, a blueprint, direction, and funding. There's cross-government coordination of that effort. That's new. That's different from the past.
Lewis described Police AI's approach: stop doing everything. Do fewer things, better. They've published something called the Joint Operating Procedure - a standardised way for CPS and policing to consider technology use together on a national board that meets every couple of months. What's the performance of that tool? What does it mean for victims and survivors? What does it mean for suspects? Having that consistency and ownership is the only way forward.
The strategy isn't about fixing everything nationally. It's about building approved, modular components. A translation component. A summarisation component. Forces can choose their own platform or provider, but if they can't get that standardised component from Police AI, there's no point them developing it themselves. They save time and effort.
"That's the sort of standardisation that I think helps initially stimulate some of the reform and broader reform. Living through doing, not necessarily setting a load of strategy on how it could happen," Lewis said.
Dave asked: "If you could remove one constraint tomorrow - policy, tech, or process - what would you change?"
Craig responded by challenging the premise. He's less interested in removing constraints per se. Rather, he wants to ensure appropriate and proportionate structure. He wants to solidify the collaboration that's starting to happen. Structured accountability and system governance. Structured funding that forces and motivates prioritisation. An Office for Criminal Justice Reform, Mark 2, maybe.
Matt said you have to kill some sacred cows. Postcode lottery is bad - but local agency is important. Mayors and Police and Crime Commissioners need local autonomy. That tension has to be addressed. And you probably need to think about it like a private business - current structure isn't sustainable. Return on investment compared to 25 years ago is poor. Change is needed.
Can AI help us experiment faster? Can we model solutions, test them, get cross-agency buy-in?
Matt talked about using AI tools for deep research, synthetic analysis, comparing what works in different places. Use case comparison could show which approaches get better outcomes, then replicate them.
Lewis highlighted that while fragmentation is a blocker, AI can help. Standardising components. Having approved tools. Creating consistency in standards. That's what Police AI is trying to do - use innovation as a way to stimulate broader reform.
But it all comes back to one thing: this is about leadership, not technology. The system can't afford to stay as it is. Change has to happen. The tools exist. The will has to come from people willing to think systemically rather than just protecting their own domain.