Artificial intelligence is already inside your agency, often before the agency has formally approved it, purchased new software, issued a policy or scheduled training.
Employees have found it on their own. An officer may use AI to improve a report; a detective may use it to organize investigative notes; a public information officer may draft a news release; an analyst may summarize a lengthy document; and a supervisor may use it to prepare a presentation.
There may be no purchase order, implementation plan or executive briefing.
That is what makes generative AI different from the major technology shifts law enforcement has managed over the past three decades.
| RELATED: Governing AI in policing: What law enforcement leaders need to know
When agencies adopted computer-aided dispatch systems, records management systems, body-worn cameras or license plate readers, leadership controlled the process. Technology was evaluated, purchased, installed, tested and governed, and employees were trained before it was widely used.
Generative AI has taken the opposite path. It entered agencies through web browsers, mobile devices and free online tools. In many cases, employees began using it before leadership had an opportunity to evaluate the risks or establish expectations.
For police leaders, the question is no longer whether AI should be allowed. In many agencies, employees have already answered that question through their day-to-day use of the technology.
The more pressing question is whether leadership will take ownership of how AI is used moving forward.
Recent research conducted through the Public Sector AI Communications Research Initiative (PSACRI) found that agencies across government face a common challenge: Employee adoption of AI is advancing faster than organizational readiness. Employees are experimenting with AI while governance, policy, training and executive oversight struggle to keep pace.
That gap creates operational, legal and reputational risks that agencies cannot ignore.
Fortunately, chiefs do not need to become AI experts overnight. They need to lead their organizations through the same disciplined process they have used to manage other significant operational changes.
Here are six places to start.
1. Assume AI is already inside your agency
Many executives continue to ask whether employees are using artificial intelligence. A better starting point is to assume that they are.
Most employees are not trying to violate policy or create unnecessary risk. They are looking for ways to work more efficiently, improve writing, organize information, summarize documents or save time on routine administrative tasks.
The issue is not employee curiosity. It is unmanaged use.
Once leaders assume AI is already being used, the conversation shifts from debating adoption to understanding current practices and establishing appropriate safeguards. That allows agencies to get ahead of the issue instead of reacting after something goes wrong.
2. Find out how AI is actually being used
Before writing an extensive policy, leaders need to understand what is already happening.
Ask straightforward questions:
- Who is using AI?
- Which tools are employees using?
- What tasks are they completing?
- Are employees using public AI platforms?
- What information are they entering into those platforms?
- Are supervisors reviewing AI-assisted work?
These conversations should not feel like an internal investigation. Employees need confidence that leadership wants to understand current practices, not punish innovation.
Many agencies will discover that employees have already identified legitimate uses that improve efficiency without increasing organizational risk. Others may uncover practices that require immediate attention. Either outcome provides valuable information.
You cannot manage what you do not understand.
3. Establish data boundaries immediately
One of the fastest ways to reduce organizational risk is to establish clear rules about what information should never be entered into public AI platforms.
That guidance should cover criminal justice information, personally identifiable information, active investigative details, juvenile records, personnel information, medical information, intelligence products, victim information and any data protected by law or agency policy.
These expectations do not require a lengthy policy manual. Simple guidance issued today is more valuable than comprehensive guidance delivered six months from now.
Employees should understand that public AI tools are not appropriate places to process sensitive law enforcement information. That single step can reduce unnecessary exposure while leaders build a broader governance strategy.
4. Keep humans responsible for every decision
Artificial intelligence can assist with work, but it should never replace professional judgment.
AI can draft reports, summarize documents, organize information, suggest edits and generate ideas. It can also produce inaccurate information, misunderstand context, omit important facts or present incorrect information with complete confidence.
The responsibility for accuracy never belongs to the technology. It belongs to the employee using it.
An officer who uses AI to improve a report still owns that report. A detective who uses AI to organize investigative notes remains responsible for the investigation. A public information officer who drafts a release with AI remains responsible for every fact released to the public. Supervisors remain responsible for every document they approve.
AI can assist the process, but it cannot assume accountability.
5. Make AI a leadership responsibility
One of the greatest risks surrounding AI is assuming someone else owns it.
Information technology manages systems and security. Legal advisors understand liability. Training personnel develop employee education. Public information officers understand public trust and community messaging. Investigators understand case integrity.
Each discipline has an important role, but without executive leadership, responsibility can become fragmented.
Chiefs should designate an individual or working group to coordinate AI governance across the organization. That group should evaluate emerging technologies, recommend interim guidance, identify training needs, monitor legal developments and provide practical recommendations to executive leadership.
AI should not become another committee project. It should be treated as a leadership responsibility.
6. Train before the first problem becomes public
Every major incident creates a learning opportunity, but good leaders try to learn before the incident occurs. Artificial intelligence should be treated no differently.
Employees need practical guidance. They should understand what AI does well, where it struggles, what information should never be entered into AI systems, how to verify AI-generated content and when human review is required.
Training should also reinforce a simple principle: Speed never outweighs accuracy.
This is particularly important for public information officers, investigators, supervisors and command staff, whose work directly affects public trust and organizational credibility.
The goal is not to make every employee an AI expert. It is to ensure every employee understands responsible use.
Leadership will determine the outcome
Artificial intelligence is already changing how employees write reports, communicate, conduct research, organize information and complete routine tasks.
Ignoring that reality will not slow adoption. It will simply leave agencies without clear expectations, safeguards or accountability.
The agencies that navigate AI successfully will not necessarily be those with the largest budgets or the newest technology. They will be the agencies whose leaders establish clear expectations, protect sensitive information, require human accountability, invest in employee training and continually evaluate how AI supports the mission.
Police leaders have successfully guided their organizations through decades of technological change. Artificial intelligence presents the next leadership challenge, but it arrived before many organizations had the opportunity to prepare.
Chiefs do not need every answer today. They do need to begin asking better questions.
Agencies that wait until an AI mistake becomes a headline will be responding under pressure rather than leading with purpose. Your officers may already be using AI. Now is the time for leadership to catch up.
| WATCH: If AI gives police time back, what should agencies do with it?