The most important question facing police chiefs is not, “Which AI tool should we buy?” It is, “Are we building the capability to solve our own problems?”
AI is entering policing through report-writing systems, license plate readers, evidence platforms, dashboards, workflow automation and analytical tools. I believe its greatest impact may ultimately occur somewhere less obvious: in the chief’s office.
The chief of 2030 will increasingly ask questions across crime, staffing, calls for service, investigations, budgets and organizational performance, receiving information assembled from multiple agency systems. That future requires something many agencies have not yet built: intentional information architecture.
Chiefs must begin thinking like data architects. That does not mean learning to write code. It means understanding where agency information resides, how systems communicate, who controls the data and how technology can be designed around the agency’s mission rather than forcing the agency to operate around a vendor’s product.
Police technology has traditionally been vendor-led: Identify a problem, watch a demonstration, purchase another application, database, login and contract. AI gives chiefs an opportunity to rethink that model and ask a better question before buying anything new: Can we securely build this capability within the enterprise environment we already own?
Many agencies already operate within enterprise environments such as Microsoft 365. Depending on licensing and configuration, those environments may include Power BI, Power Apps, Power Automate, SharePoint, scripting and automation capabilities, cloud services and AI tools.
Yet agencies may purchase separate commercial products to perform functions that could potentially be built, integrated or automated using technology they already license. That does not mean police departments should become software companies. It means leaders should first inventory existing capabilities and determine whether customized workflows, dashboards, applications or integrations can solve the problem securely, sustainably and at a reasonable total cost over time.
Design technology around the agency
One size does not fit all in policing. The needs of a 12-officer department differ from those of a 3,000-officer metropolitan agency. Policing an airport presents different challenges than policing a residential community, and a tourist destination operates differently from a university, transit system or rural county.
Our technology architecture should reflect those differences.
This is where chief AI officer (CAIO) capability becomes important. I am not suggesting every department needs another full-time executive. Most could not afford one. But agencies increasingly need access to CAIO-level expertise that connects operations, data architecture, cybersecurity, contracts, automation, AI governance and organizational strategy into one coherent framework for leadership decisions.
For smaller agencies, that capability might be fractional, regional or shared.
A CAIO can help a chief inventory what the agency already owns before purchasing something new:
- What systems are we already paying for?
- Where does our data reside?
- Which systems can communicate through secure application programming interfaces (APIs)?
- Where can workflows eliminate repetitive processing? Can existing enterprise tools solve the problem?
- Where can AI help leaders analyze information?
- What permissions should those systems have?
- Where might emerging connection standards, such as Model Context Protocol (MCP), appropriately fit?
- What should we build, and what truly requires a specialized vendor?
The objective is agency-specific decision architecture.
Build, buy and prove value
Using existing enterprise platforms can sometimes eliminate redundant products and avoid substantial vendor markups. Building, however, is not automatically cheaper. Agencies must consider cybersecurity, maintenance, technical expertise, continuity, records requirements and long-term support.
The point is simple: “Buy another product” should no longer be the automatic answer.
When an outside product is necessary, chiefs should insist on proving its value in their environment. Vendor demonstrations show what a system can do under conditions the vendor controls. Chiefs need to know what it does with their data, workflows, personnel and error tolerance before committing the agency to broad deployment or creating new dependencies.
Establish a baseline and conduct a limited pilot. Ask what outcome should improve, what constitutes an unacceptable error, whether personnel can identify and correct mistakes, and how the agency will measure accuracy and reliability under actual operating conditions.
Leaders should also determine whether the system integrates with existing technology, whether information is accessible through documented APIs, whether data can be exported when the relationship ends and what conditions would trigger suspension or termination.
Ultimately, chiefs should be able to answer one deceptively simple question: Can we explain, verify, audit and exit this system?
If not, the agency may not own a capability. It may simply have created another dependency that becomes harder and costlier to unwind.
Redesign work, not accountability
Chiefs should also be cautious about promises that AI will simply reduce workload.
AI can automate transcription, retrieval, routing, summarization, document assembly and information gathering. But the work does not disappear. It moves.
Officers may spend less time assembling information while analysts spend more time verifying outputs. Supervisors may spend less time searching but more time evaluating exceptions. Records personnel may manage new audit trails. Attorneys may confront new discovery questions. Leaders may suddenly have access to information they never had before.
That is not necessarily failure. It is organizational redesign, requiring leaders to decide where recovered capacity should be deliberately reinvested for greater operational impact.
Human oversight must also be structured, not symbolic. Humans do not need to manually approve every automated step. A workflow can retrieve information. Power Automate can route it. A dashboard can aggregate it. AI can summarize it. An AI agent may help a commander analyze information across multiple systems and identify patterns or questions the leader had not considered.
The human belongs at the judgment point.
The greater the consequence, the stronger the oversight should become. Automate collection, routing and repetitive processing. But when decisions affect liberty, credibility, discipline, surveillance or force, a trained person must understand the information, assess its reliability and own the decision.
Never automate accountability.
Control the data — and the exit
Data control may be the issue chiefs overlook most.
A contract can say the department owns its data while a vendor hosts, queries, backs up and controls access to it. Chiefs should ask who can use the data, who maintains audit logs, who reports a breach, who bears responsibility when something goes wrong, how the technology affects the agency’s cybersecurity and insurance obligations, and whether the agency can leave without losing operational continuity.
These questions extend beyond the real-time crime center (RTCC). Real-time policing is a decision-making capability, not merely a room of monitors.
As AI becomes more deeply integrated into police operations, chiefs must ensure that technology strengthens their agency’s ability to make informed, accountable decisions rather than creating new dependencies they cannot fully understand or control.
The goal is not to acquire more AI. It is to build an agency that knows what problems it needs to solve, what capabilities it already possesses and how to determine whether new technology genuinely improves performance.
I welcome the opportunity to help departments navigate these challenges and discuss how they can build AI capabilities that support their operational needs. Police leaders interested in continuing the conversation can reach me through Police1.
