Frontline policing, OSINT, and the gap between policy and practice

Open-Source Intelligence (OSINT) is a rich source of information for UK policing, but translating it into frontline practice remains a challenge for many officers. Some are unsure what is permitted under policy, while others simply do not know what tools are available to them. As part of the Police Now Graduate Scheme, Principle One hosted three officers from Safer Neighbourhood Teams (SNT) from the Metropolitan Police Service (MPS) for a four-week secondment to explore why this gap exists and what could realistically be done about it.

The wards they police could not have been more different. Ali is a Dedicated Ward Officer in East London; a ward with green open spaces and a residential community shaped by it. Stephen is based in South London, where the town centre carries the pressures of high-density outer London policing with all the key demands that a busy transport and retail hub generates. Finally, Georgie’s ward in central London is shaped by retail, tourism, and embassies, and the policing reality reflects that. Despite those differences in their neighbourhood roles, they found similar problems in the challenges each faced with the secondment representing a valuable opportunity to take a more holistic approach to problem solving.

Policing meets consultancy

Becoming consultants for the duration of their secondment was a culture shock in many ways. Twenty months into policing, the most senior officer most of them had interacted with was an Inspector. “The MPS has its own ways of marking authority, and this has become second nature”, Stephen reflected. “Walking into a building where rank simply wasn’t part of the conversation was a positive surprise.”

Right from the start, they were introduced to the language and practices of consultancy: agile working, sprint cycles, and creating artefacts. Adjusting to the shift in culture took time. Weekly end-of-sprint reviews were attended not just by the project team but also by senior stakeholders within Principle One who would ask direct, useful questions and expect direct, useful answers. “At first this was intimidating but it quickly became a valuable opportunity to sense-check our reasoning and approach.”

This reflected a broader difference between policing and consultancy. Both are fundamentally about problem-solving, but in policing the work is often reactive – responding to incidents and managing immediate risks. Consultancy, however, offers the space to step back, analyse root causes, and test assumptions before committing to a course of action. Artefacts are produced to documented not just what was decided, but why. While the MPS trains officers in models such as SARA (scanning, analysis, response, assessment), the secondees found that consultancy methods enabled them to apply these frameworks in a meaningful way and arrive at more effective solutions.

Building Ossist

The secondment brief looked straightforward but was far more complex in practice: Examine how Open-Source Intelligence (OSINT) is used by frontline officers, identify why it is not used as well as it could be, and propose a solution to address this.

Engagement with Safer Neighbourhood Teams and other stakeholders helped identify the uncertainty around what is and isn’t acceptable around OSINT under MPS policy with officers uncertain what they can and can’t do on MPS devices.

The work quickly found its focal point from there. Through engagement with the MPS Internet, Intelligence and Investigation (i3) Team, with national policing leads, subject matter experts, and with serving officers across multiple boroughs, the problem statement was distilled to this:

OSINT in frontline policing is underused – this is driven by a lack of awareness and confidence in the existing tools, and the current practice of officers carries both personal safety and legal risk.

Three potential solutions were identified: a guidance tool in the form of a decision tree or chatbot; an SNT SPOC network recommendation; and a policy review.

The choice of which solution to pursue came down to one question: what does an officer need when they are working on an investigation under pressure and with no time to read a lengthy policy document? The solution was a chatbot which was named Ossist (Open Source + Assistant), built on Microsoft Copilot Studio and grounded in a knowledge base constructed by the secondees themselves. Ossist is designed to give a frontline officer immediate answers about legislation, tools and authorisation pathways in the moment they are needed.

Because the secondees were also the intended users, they found themselves constantly testing the idea against the realities of frontline policing. “Would I trust this tool on a Safer Streets operation? Would my colleagues even know when to use it?” Some of the most useful parts of the project came from challenging the tool from the perspective of the officers it was designed to support.

A working prototype was built in three days. None of the secondees had previous technical experience and they expected this to limit the technical scope of the project. However, they found Microsoft Copilot Studio easy to work with and made rapid progress to incorporate the features they wanted into the proof of concept.

The difficult part was everything surrounding the build. While working on the proof of concept, it wasn’t possible to get direct access to MPS OSINT policy and guidelines, making it challenging to build a credible knowledge base that could handle specific frontline scenarios. Instead, learning from Principle One’s synthetic data capability, the secondees used generative AI tools to create synthetic policy documents that mirrored the structure and language of the originals – enough for an initial proof of concept.

Next there is the problem of how a solution like Ossist would work in the long run. Turning the proof-of-concept into something that could be deployed on officers’ devices and be trusted for day-to-day operational use is a different challenge entirely. Information assurance, data access and quality issues, Data Protection and licensing considerations, and significant stakeholder engagement would all be needed to roll out a live product.

Ali reflected that “recalibrating what good looks like, from ‘produce a finished solution’ to ‘build the foundation for the right product to follow’ was one of the harder lessons we have learned over the last four weeks”. Despite that, they achieved a lot within a short timeframe: a working prototype that proves the concept; an evidence base that justifies further investment; and a much clearer understanding of the scale of work that real deployment would require.

Bridging the gap between policy and practice

Crime increasingly leaves a digital footprint, even in neighbourhood policing. Officers are expected to make quick decisions involving online material while balancing legislation, force policy and public expectations, often with limited guidance available in the moment. Over the four weeks, the secondees realised that the issue wasn’t whether frontline officers wanted to use OSINT. Most already were, the problem was that many were doing so without confidence or consistency, and sometimes without knowing whether its use was supported by policy.

Our OSINT experiment found that while AI tools like Ossist cannot replace the judgement, experience, or local knowledge of a Safer Neighbourhood officer, what they can do is lower the barrier for officers trying to do the right thing while up against the day-to-day pressures of a frontline role.

Principle One and Police Now would like to thank the MPS i3 Open Source Unit, including DI Dave Batt and PS Eleanor James, and DI Claire Harvey and Inspector Liam Cahill for all their support throughout the secondment.