Unique and innovative ideas for solving policing challenges were presented by members of the police workforce at the annual conference of the Police Superintendents’ Association (PSA). 

The event, held from the 14-15th September in Stratford Upon Avon, brought together senior leaders in policing to discuss ‘Shaping the Future: Innovation and leadership in Police Reform’. 

As part of the event, finalists from the Flashlight Innovation Project presented their ideas for new solutions to policing challenges. 

Earlier this year, the PSA launched a new collaboration with Amazon Web Services (AWS), Society of Evidence Based Policing (SEBP), Accenture and the Open University Centre for Policing Research and Learning (CRPL), asking members of the policing to pitch ideas for innovative technology to support policing.  
 
Nominations were evaluated by a judging panel consisting of PSA representatives and external experts. Five shortlisted projects were then supported in developing their ideas into proof of concepts, before being invited to present on their innovation at the conference, to an audience of peers and policing leaders.  

Each finalist presented their concept with test platforms to demonstrate their solution, before the judging panel were asked to select a winner, alongside an audience vote. 

The winning pitch came from ACC Lee Berry of the National Police Air Service and Chief Inspector Dan Tillett of West Yorkshire Police.

Their innovation was Project GERDA: Accelerating child victim identification through responsible technology. 
 
GERDA was inspired by West Yorkshire Police investigations in which children subject to online abuse had to be identified manually from images. Extensive partnership working and a subsequent controlled facial recognition exercise demonstrated that technology could reveal safeguarding and investigative connections that established processes had not. It also reinforced the need for trained human review, and decision making, strong governance and secure information handling. Developed with policing, safeguarding and technology partners, GERDA is a proof of concept designed to provide a controlled pathway from an unidentified image to a potential safeguarding lead. 

PSA Vice President Sara Crane said: “It was fantastic to launch this project with the wider team and to host such an engaging and interesting session at conference. 

“We were overwhelmed by the quality and quantity of applications we received. They demonstrated such incredible innovation and problem-solving, utilising developing technology, from people across policing who are so clearly passionate about making our processes and ways of working better for the service and for communities. 

“It was a very difficult job to select our five finalists, but they were all extremely impressive and presented brilliantly to our conference delegates. 

“I’d like to thank each of them for their efforts in creating their innovations in force, working with the team at Accenture in refining their solutions, and then presenting so well on what are technical and complex projects.  

“Congratulations to Lee and Dan, who will now receive further support from technical experts in AWS and Accenture in further developing their concepts. We look forward to seeing their idea progress.” 

All finalists will remain in conversation with the Flashlight technical experts to further explore ways to develop their ideas. 

Other shortlisted projects, which were expertly presented to judges were: 

Detective Constable Andrew Cartwright, North East Regional Organised Crime Unit 
Detective Inspector Tony Forster, Cyber, North East Regional Organised Crime Unit

A murder investigation gets HOLMES, a major incident room and a team to read every statement. A cash-in-transit robbery gets none of that. It can still run to dozens of statements from guards, shop staff and passers-by, held by one officer alongside a full workload. When one witness describes "the one who grabbed the box" and another "the man in the grey balaclava", the connection is only made if somebody reads both and remembers. That is where things get missed. 

TRACE lets that officer ask a question in plain English and get an answer in minutes. It uses retrieval-augmented AI to search an entire set of MG11 statements by meaning rather than keyword, surfacing relevant passages however differently each account described them. Every answer cites the statement, witness and paragraph it came from, so it can be checked against the original. 

TRACE does not draw conclusions. It finds what is already in the evidence. The investigator decides what it means. 

TRACE is being built with Accenture. It is evaluated entirely on synthetic statements, with no operational material, against a locked ground-truth key measuring precision, recall, citation accuracy and hallucination rate. 

A/Detective Inspector Raymond Sekalongo, Metropolitan Police 

This project proposes a national, technology-enabled biometric intelligence exchange system to improve the detection of cross-force offending and Serious Organised Crime (SOC). Current forensic intelligence sharing is fragmented, with biometric intelligence-only and full identifications often remaining within local force systems, limiting visibility of travelling offenders, linked crime series and wider SOC networks. 

The proposed system would securely receive biometric identifications from all forces, standardise and share them in real time, and cross-reference them with SOC intelligence datasets. Using analytics, machine learning, geospatial and network analysis, the platform would identify patterns, generate risk scores and produce actionable intelligence packages for investigators. 

The system is designed to be scalable across all forces, legally compliant, and aligned with NPCC priorities around data sharing, digital evidence exploitation, analytics capability and innovation. It would reduce duplicated effort, improve investigative efficiency, support earlier SOC disruption, strengthen OCG mapping and enhance safeguarding by identifying exploitation patterns sooner. 

Suzie Kettridge, Senior Technologist, Metropolitan Police 

This project proposes an AI-enabled image analysis platform for policing that transforms large volumes of unstructured image data into searchable, actionable intelligence. Designed to operate securely in either cloud or on-premises environments, is designed to enable processing of thousands of images at once, extracting embedded text through OCR, detecting languages, capturing metadata, and automatically categorising visual content such as people, vehicles, weapons, documents, and locations. 

Investigators can use natural language and keyword searches to quickly identify relevant material, filter images by category, and uncover patterns across datasets. The system would incorporate configurable risk-based prioritisation, explainable AI features, audit trails, geospatial mapping, and exportable reporting. It would also support evolving criminal slang and terminology through user-led updates. 

Developed in three phases, the solution would deliver OCR and search capabilities, AI-powered image classification, and advanced analytics including summarisation, multilingual search, relationship discovery, and potential video analysis. The project addresses the growing challenge of reviewing vast amounts of digital evidence, helping investigators prioritise high-value material, reduce review times and costs, improve decision-making, and minimise unnecessary examination of personal data while maintaining human oversight. 

Supt Alex House, Bedfordshire Police 
Dan James, Project / Programme Manager, Bedfordshire Police

Project Daedalus recombines AI with the well-established and effective SARA Problem-solving model to modernise problem-oriented policing supporting the scanning, analysis, response and assessment phases.  

This solution will use data connections and policing information to bring relevant material into one workflow. This will assist officers and analysts by accelerating their ability to efficiently identify recurring crime, disorder and vulnerability issues, understand their underlying causes, develop targeted responses, and assess whether those responses are working.  

AI will assist with data gathering, quality checks, pattern identification and evidence-based critical analysis, while officers and analysts retain oversight and professional judgement as the humans in the loop at each stage.  

The tool will provide a shared platform for risk assessment, collaboration, escalation, review and learning, ensuring plans remain live, visible and adaptable as circumstances change. For policing, this will significantly reduce time spent on station, enable quick decision making and action, improve consistency, increase the number and quality of problem-solving plans, and support better use of analyst and officer time, ultimately reducing repeat demand.  

For communities, this means faster intervention, more effective prevention activity, fewer repeat victims, reduced crime and disorder, and improved confidence that problems are being identified, understood and addressed through evidence-led policing.