CPS’s pitch — that outcome-focused AI agents built on Microsoft Copilot can shrink paperwork, speed investigations and put more officers back on the street — arrives at a moment when policing is hungry for tangible efficiency gains and painfully wary of unproven AI experiments. The company will demonstrate those claims at The Emergency Tech Show (ETS) in the Microsoft Partner Pavilion on 17–18 September 2025, promising live examples of Copilot Agents, shift-briefing automation and back-office process automations designed specifically for UK policing. (emergencytechshow.com) (emergencytechshow.com)
Microsoft’s Copilot family — embedded across Microsoft 365, Power Platform and Azure — gives partners an enterprise-grade foundation to deploy AI in a way that integrates with the tools forces already use: Teams, Outlook, SharePoint and Dataverse. Independent vendor and academic work suggests Copilot-style tools can and do yield measurable productivity gains in knowledge-work contexts, which is why systems integrators and specialist partners are now building domain-specific agents for regulated sectors like policing. (tei.forrester.com, arxiv.org)
However, several specific performance claims that appear in event previews or vendor interviews warrant caution. For example, public independent research for Copilot shows strong but not uniform ROI; the widely cited Forrester ranges (132–353% ROI) are materially different from anecdotal “800% ROI” figures sometimes quoted in vendor pitches. A comprehensive public corroboration of a force achieving an 800% ROI from a Copilot trial — or productivity savings “equivalent to six fully kitted police vehicles” attributed to a single Copilot pilot — was not locatable in independent reporting at the time of writing. If those numbers were supplied by CPS in interviews, they should be treated as vendor-provided metrics until backed by independent, force-level case studies or audited ROI reports. In short: plausible, potentially large gains exist, but extraordinary claims require verifiable public evidence. (tei.forrester.com, cps.co.uk)
That means two practical truths for police leaders and procurement teams attending ETS:
The promise is not that AI will replace judgement; it is that well-engineered agents will free time for the judgements that only humans can make, while producing auditable, repeatable and compliant outputs that support more timely justice. The proviso is significant: those outcomes only follow from rigorous governance, clear metrics and independent validation — not marketing claims alone. (tei.forrester.com, gov.uk)
For police IT leaders, digital transformation leads and command teams attending The Emergency Tech Show, the practical test is straightforward: ask for demonstrable, tenant-bound implementations with logs and measured before-and-after outcomes — and treat any exceptional ROI figures that lack public corroboration as vendor-supplied claims rather than proven fact. (emergencytechshow.com, cps.co.uk)
Source: Emergency Services Times https://emergencyservicestimes.com/2025/08/15/cps-to-demonstrate-outcome-focused-ai-for-policing-at-the-emergency-tech-show/
Background
Why policing is interested in Copilot and agent-driven automation
Police forces across the UK face the same operational reality: growing caseloads, squeezed budgets and extensive administrative overhead that pulls officers and investigators away from frontline and high-value investigative work. National reviews and force-level case studies show automation and AI can return officer hours at scale — from redaction tooling that once cut days of work to RPA bots that reclaim tens of thousands of hours across forces. Those productivity efforts are part of a broader push to modernise policing workflows and data sharing around national platforms such as NPCE (National Police Capabilities Environment). (gov.uk)Microsoft’s Copilot family — embedded across Microsoft 365, Power Platform and Azure — gives partners an enterprise-grade foundation to deploy AI in a way that integrates with the tools forces already use: Teams, Outlook, SharePoint and Dataverse. Independent vendor and academic work suggests Copilot-style tools can and do yield measurable productivity gains in knowledge-work contexts, which is why systems integrators and specialist partners are now building domain-specific agents for regulated sectors like policing. (tei.forrester.com, arxiv.org)
What CPS says it is delivering
From the marketing: outcomes, agents and governance
CPS (Corporate Project Solutions) positions itself as a Microsoft-aligned systems integrator with long experience in public sector digital transformation. On its website the company describes a shift from simple Copilot demonstrations to agent-led workflows that are “tied to outcomes” — i.e., each Copilot Agent is designed to reduce measurable waste (admin time), speed case progression or bolster compliance. CPS lists collaborations with forces and national policing bodies and promotes a suite of Copilot Agent use cases: briefing summaries for duty handovers, FOI automation for back-office teams, automated evidence-tracking and real-time workload dashboards for commanders. CPS also states it will exhibit these capabilities at ETS in the Microsoft Partner Pavilion. (cps.co.uk)The headline claims that will draw attention at ETS
- Agents that summarise incidents and produce shift-brief packs automatically, saving officers preparatory time.
- Automation of FOI and routine back-office processes, reducing hours spent on repetitive document compilation.
- Copilot-enabled case summaries, evidence indexing and triage to accelerate investigations.
- Real-time visibility for commanders via Power Platform dashboards to improve resource allocation and KPI tracking. (cps.co.uk)
What the wider evidence base says about likely impact
Measurable productivity gains are plausible — but vary by context
Third-party analyses and early enterprise studies of Copilot point to substantial but variable productivity improvements. Forrester’s commissioned TEI modelling for Microsoft found projected ROI ranges from roughly 132% to 353% over three years for typical organisations using Microsoft 365 Copilot — with benefits driven by faster document production, reduced search time and improved collaboration. An independent large-scale study of Copilot adoption also reports meaningful time savings on routine tasks across diverse firms. These independent data points support CPS’s central premise: Copilot-based assistants can free time for higher-value tasks when correctly scoped and governed. (tei.forrester.com, arxiv.org)Where policing-specific gains have been recorded
Policing-specific automation work (before Copilot’s ubiquity) has already demonstrated significant returns: national case studies show auto-redaction tools, RPA for data-cleaning and automated triage saved forces tens of thousands of officer-hours and produced measurable cost savings. Those projects offer a realistic analogue: if Copilot Agents deliver similar fidelity and integration, forces could see comparable impact in administrative and evidence-processing workflows. However, policing workloads are functionally different from office knowledge work; integration with case management systems, admissibility standards and strict audit trails are non-negotiable prerequisites for translating productivity into operational value. (gov.uk)Verification and independence: what is claimed versus what’s proven
Claims made by CPS and the independent record
CPS states it has delivered “measurable improvements” — and references partnerships with regional forces. Those are valid organisational claims and CPS is listed as an exhibitor in the Microsoft Partner Pavilion at ETS, confirming its presence at the event. (emergencytechshow.com)However, several specific performance claims that appear in event previews or vendor interviews warrant caution. For example, public independent research for Copilot shows strong but not uniform ROI; the widely cited Forrester ranges (132–353% ROI) are materially different from anecdotal “800% ROI” figures sometimes quoted in vendor pitches. A comprehensive public corroboration of a force achieving an 800% ROI from a Copilot trial — or productivity savings “equivalent to six fully kitted police vehicles” attributed to a single Copilot pilot — was not locatable in independent reporting at the time of writing. If those numbers were supplied by CPS in interviews, they should be treated as vendor-provided metrics until backed by independent, force-level case studies or audited ROI reports. In short: plausible, potentially large gains exist, but extraordinary claims require verifiable public evidence. (tei.forrester.com, cps.co.uk)
What attendees should look for at ETS (practical checklist)
- Demonstrations that use real police data schemas
- Watch how the agent connects to live case management, evidence stores and IdAM systems (Entra). Confirm it respects data boundaries and requires force credentials.
- Auditability and explainability
- Ask to see thread-level logs, action traces and decision trails; check whether outputs are linked to sources and whether there’s a human approval gate for investigative conclusions.
- Integration with existing workflows
- Validate the flow from Teams/Outlook to the force’s Records Management System or Niche; test whether agency retention policies and legal hold are preserved.
- Deployment model and tenancy
- Confirm whether the solution runs inside the force’s Azure tenant (preferred for sovereignty) or whether third-party storage is used.
- Training, change management and staff uplift
- Probe CPS for delivery of training, skills-transfer and a practical plan to embed agents into shift patterns — not just pilot demos. (cps.co.uk, emergencytechshow.com)
Strengths of CPS’s approach
- Platform alignment: Building on Microsoft 365, Power Platform and Azure lowers friction for forces already standardised on Microsoft tooling. This reduces integration risk and shortens time-to-value. (cps.co.uk)
- Outcome focus: The “tie an agent to an outcome” philosophy is pragmatic. Designing agents around measurable KPIs (hours saved; case progression times; FOI turnaround) helps forces justify investment and maintain governance discipline.
- Agent Factory and repeatability: If CPS genuinely offers an “Agent Factory” — a repeatable, secure development lifecycle for agents — it addresses a common failure mode of pilot projects: bespoke, one-off implementations that never scale.
- Operational telemetry: Using Power Platform dashboards and Copilot telemetry can give command teams real-time situational awareness of workloads, enabling dynamic resource allocation. (cps.co.uk)
Risks, governance and the all-important caveats
1) Data protection and legal admissibility
Police use cases require strict chain-of-evidence and legally defensible documentation. Any agent that summarises witness statements or body-worn video must preserve provenance, ensure summaries can be audited back to source material and never introduce untraceable edits. Failure to enforce those controls risks evidence being ruled inadmissible or complaints about investigative integrity.2) Hallucinations and inappropriate automation
Generative AI can produce confident but incorrect outputs. In policing, an incorrect summary or misattributed fact is more than an inconvenience — it is operationally harmful. Agents must be deployed with deterministic retrieval-augmented generation, conservative refusal modes for low-confidence outputs and mandatory human sign-off on any investigatory claims.3) Shadow IT and unmanaged AI adoption
One driver for vendor-led Copilot solutions is the desire to offer a sanctioned alternative to staff turning to consumer-grade AI tools. But if governance is weak, teams will still adopt unsanctioned tools. Effective deployment must pair secure, sanctioned agents with strong policies, tenant-level controls (Purview, DLP) and continuous education.4) Over-reliance and deskilling
Automation that removes routine tasks can free time — but long-term over-reliance can erode skills. Police forces must balance efficiency efforts with skills-retention and training programs; automation should augment, not replace, painstaking investigative craft.5) Vendor lock-in and platform dependency
CPS’s approach — deep Microsoft integration — reduces short-term integration work but increases platform dependency. Forces must insist on exportable artefacts, documented APIs and portability plans in procurement to avoid high switching costs later. (tei.forrester.com)Practical governance checklist for commissioners and police CIOs
- Require an initial short prototype (4–8 weeks) with a narrow, high-value use case, then a controlled production roll (3–6 months) under SLAs.
- Mandate detailed observability: per-thread decision logs, timestamped data-access records, and model versioning for every agent action.
- Enforce identity and least-privilege execution: agents should act with the same or lower privilege than the initiating officer and respect Microsoft Entra role assignments.
- Insist on tenant-resident deployments: solutions should operate inside the force’s Azure tenant or NPCE environment where feasible.
- Demand a rollback plan and an export path for data and configurations; ensure contractual audit rights and external validation clauses.
- Build staff-focused measures: time-savings per role, change-management milestones and retention/training KPIs. (cps.co.uk)
Where CPS fits in the broader UK policing AI ecosystem
CPS is one of many systems integrators racing to operationalise Copilot and agent strategies for public sector customers. The Microsoft Partner Pavilion at ETS consolidates that ecosystem, making it easier for forces to compare vendors and see demonstrable outcomes rather than slides. The existence of a national platform agenda — including NPCE and the Police Assured Landing Zone designs — means forces now have options for assured cloud tenancy, which is crucial for scale adoption of agentic AI across policing. (emergencytechshow.com)Bottom line: realistic opportunity, avoidable risks
CPS’s approach is the most useful kind of AI sales pitch: operational, outcome-oriented and anchored to the Microsoft stack that most UK forces already use. Independent research on Copilot’s productivity impact is encouraging and supports the idea that task automation and retrieval-augmented summarisation can free substantial officer time when implemented with discipline. For policing, where evidence integrity and governance are paramount, the differentiator will not be the agent’s creativity but its auditability, tenancy model and integration with legal-grade workflows. (tei.forrester.com, gov.uk)That means two practical truths for police leaders and procurement teams attending ETS:
- Validate measurable claims with documented, auditable case studies (force-level ROI analyses, before-and-after time-motion studies).
- Treat governance as the product’s primary feature — require per-action logs, model/version controls, tenant residency and human-in-the-loop fail-safes.
Final assessment — why this matters for frontline policing
If properly scoped and governed, Copilot Agents can move policing away from repetitive paperwork and towards operational intelligence — not speculative predictive policing, but practical decision support that surfaces risk, prioritises caseloads and helps commanders allocate scarce resources more effectively.The promise is not that AI will replace judgement; it is that well-engineered agents will free time for the judgements that only humans can make, while producing auditable, repeatable and compliant outputs that support more timely justice. The proviso is significant: those outcomes only follow from rigorous governance, clear metrics and independent validation — not marketing claims alone. (tei.forrester.com, gov.uk)
For police IT leaders, digital transformation leads and command teams attending The Emergency Tech Show, the practical test is straightforward: ask for demonstrable, tenant-bound implementations with logs and measured before-and-after outcomes — and treat any exceptional ROI figures that lack public corroboration as vendor-supplied claims rather than proven fact. (emergencytechshow.com, cps.co.uk)
Source: Emergency Services Times https://emergencyservicestimes.com/2025/08/15/cps-to-demonstrate-outcome-focused-ai-for-policing-at-the-emergency-tech-show/
