The wrong vendor doesn’t just waste your time—it exposes your agency to real risk.

AI vendors are showing up in inboxes, conferences, and council meetings. Many offer tools that could help your agency—transcription, analytics, dispatch optimization. But the reality is this: most vendors are not prepared to answer the questions that matter to command staff, legal, or IT.

If you schedule a demo before screening for risk, you’re giving them your time—and possibly your credibility—without doing your job.

‍

What to Ask Before the Demo

1. Does the system make operational decisions or trigger responses?

If the tool influences resource deployment, public messaging, or enforcement flags—it’s not low risk. These systems require governance, HITL review, and public disclosure. If the vendor says otherwise, that’s a red flag.

2. Can they show audit logs, override controls, and version history?

Every AI tool used in law enforcement must be traceable. You should be able to see who used it, what the output was, and how it was acted on. No logs? No transparency.

3. Has the system been tested for bias, misuse, or failure modes?

If the vendor can’t show results from red teaming, adversarial testing, or basic bias analysis—they’re not enterprise-ready. These are table stakes in 2025.

4. Will they contractually support audit rights, kill switches, and public disclosure?

This is where most vendors balk. If they won’t agree to suspend the tool during an investigation—or disclose basic system safeguards to the public—don’t move forward. You can’t afford to deploy tools you can’t shut off or defend in court.

‍