AI Assistant ‘Cecil’ Could Save Police 23,000 Hours a Year, Trial Finds

TL;DR: Humberside Police and Coeus Software’s “Project Moriarty” trial demonstrates that AI-powered assistant Cecil could save over 23,000 officer hours annually. Officers rated the system 4.6/5 for accuracy, with response times averaging 5-20 seconds for procedural guidance.

Humberside Police and Coeus Software have released findings from “Project Moriarty,” a trial demonstrating how an AI-powered assistant named Cecil could significantly enhance frontline policing operations. The initiative was funded by the National Police Chiefs’ Council and National Science and Innovation Board.

How Cecil Works

Cecil functions as a digital mentor within Coeus Software’s PoliceBox mobile platform. Officers can pose conversational questions about procedural scenarios—such as “I’ve arrived at a road traffic accident”—and receive relevant guidance within seconds.

The system uses a Retrieval-Augmented Generation (RAG) AI engine to analyse authorised police materials without relying on external data sources, ensuring responses remain grounded in official police procedures and policies.

Trial Results

Officers rated Cecil highly across multiple dimensions:

  • Accuracy: 4.6/5 (92% rated 4 or higher)
  • Relevance: 4.5/5 (88% rated 4 or higher)
  • Ease of use: 4.6/5 (84% rated 4 or higher)
  • Usefulness: 4.4/5 (86% rated 4 or higher)

Response times averaged between five and twenty seconds, with half of queries answered within seven seconds—a critical timeframe for frontline decision-making.

Impact Projection

If deployed force-wide, Humberside Police estimates Cecil could save over 23,000 officer hours annually—equivalent to returning several full-time personnel to frontline duties. This efficiency gain comes not from reducing headcount but from enabling officers to spend more time on policing activities rather than searching for procedural information.

Addressing the Experience Gap

The initiative addresses a critical staffing challenge: as of March 2024, 35% of police officers in England and Wales had fewer than five years’ service, compared to 14% in 2016. This dramatic shift means a significant portion of the force lacks the institutional knowledge that traditionally came from working alongside experienced colleagues.

Cecil aims to bridge this experience gap by providing instant procedural guidance to newer officers, effectively democratising access to procedural knowledge that would otherwise require years of experience to accumulate.

Looking Forward

The success of Project Moriarty suggests practical applications for AI in public services where rapid access to accurate procedural information can improve outcomes. The RAG approach—using AI to retrieve and synthesise information from authorised sources rather than generating responses from general training data—appears particularly well-suited to contexts requiring high accuracy and compliance with established procedures.

The project’s focus on measurable outcomes (hours saved, response times, user satisfaction ratings) provides a model for evaluating AI implementations in other public sector contexts.

Source Attribution:

Share this article