Government Procurement Playbook: AI Pilots for Public Safety Use Cases

A playbook emerges to pilot AI and emerging technologies for mission critical operations.

Public safety leaders have no shortage of ideas—the real challenges are a lack of time, clean data, and risk-managed ways to test new technology without losing political capital. But there's good news: a proven playbook is emerging from cities that are successfully navigating this process.

This playbook offers pragmatic steps any jurisdiction can take to partner with startups on AI and other emerging technology for public safety. It also provides a regional model that smaller agencies can adopt.

Here’s how to get started:

1. Define the Problem, Not the Product 

Before you consider any specific technology, first map the operational pain point you want to solve, such as faster incident triage, better missing-person recovery, or fewer false alarms. Then, publish brief "problem briefs" that describe the desired outcome, rather than prescribing the tool you believe will achieve it.

This approach aligns with recent guidance from the National Academies, which advises state and local governments to know their users, define the purpose, and be cautious by piloting, measuring, and iterating before scaling.

How to do it 

• Create a one-pager for each problem, detailing the who, what, when, success metrics, and necessary data.

• Invite startups to propose their own approaches; avoid rigid, prescriptive specifications in the initial stages.

• Require simple logic models that show inputs, activities, outcomes, and equity impact.

2. Build Lightweight Guardrails Before You Pilot 

The most successful cities are the ones that establish basic governance upfront. This includes risk triage (classifying use cases as low, medium, or high risk), clear expectations for human oversight, data use rules, and a plan to measure accuracy and harm.

For example, San José evaluates AI use cases based on risk and purpose, and then pilots them with monitoring. This "experiment, measure, adapt" rhythm is exactly what the National Academies recommends.

Starter Guardrails 

• Risk Triage: Classify proposals based on data sensitivity, public impact, and the level of automation.

• Human Oversight: Define when a human must approve, review, or override the technology.

• Measurement Plan: Agree in advance on metrics, sampling, bias checks, and exit criteria.

3. Invest in People and Data 

Every successful AI initiative is built on a foundation of data quality and staff capability. Local governments that are advancing fastest treat digital equity and data as foundational elements. Better connectivity and data literacy not only unlock trustworthy AI but also reduce the risk of widening digital divides. The goal is to get your "data house in order" first, and then build AI on top.

Quick Wins 

• Establish a cross-departmental data governance huddle that includes IT, legal, public safety, and HR.

• Launch basic data literacy training for command staff and add role-specific AI training.

• Link all AI pilots to your broadband and digital equity plan to ensure the benefits reach everyone.

4. Copy San José’s “Civic Sandbox” Model 

San José has actively lowered the barriers for startups. The city recently awarded three $50,000 grants and one $25,000 grant to early-stage AI companies. These grants are paired with professional support (such as real estate, legal, and IT consulting) and ongoing meetups to keep the tech ecosystem connected.

The city also upskills its own staff through a 10-week AI program, ensuring internal capacity grows alongside the pilot projects. Other jurisdictions can adapt this "incentive + support + skills" approach without a massive budget.

What You Can Borrow 

• A micro-grant program tied to clear ethical AI standards and community benefit.

• A standing meetup for government leaders, vendors, and academics to match problems with solutions.

• A simple intake form to route AI ideas through your risk triage process.

5. Use a Regional Model if You’re a Smaller Agency 

If your agency lacks the scale of a city like San José, consider joining forces regionally. In Greater Phoenix, for instance, a nonprofit consortium called The Connective helps cities pool resources, define common problems, run calls for innovation, and deploy technology across multiple jurisdictions.

The Connective is housed within the Partnership for Economic Innovation, a 501(c)(3) organization. This provides the effort with neutral governance and the ability to braid funding and expertise.

How a Regional Consortium Helps 

• Shared Capacity: It centralizes resources like legal and procurement templates, and data governance.

• Better Pilots: You can test a technology once, evaluate it rigorously, and then replicate it across multiple agencies.

• Vendor Accountability: It provides one common standards framework for many different buyers.

6. Make Procurement Work for Pilots 

Traditional RFPs are often too cumbersome for a learning process. Instead, use staged procurement:

• Discovery Sprints: Short, 2-to-6-week sprints to validate data access and feasibility.

• Paid Pilots: 60-to-120-day pilots with explicit metrics and a "go/no-go" decision gate.

• Scale Contracts: Contracts that are contingent on hitting pre-defined safety and performance thresholds.

San José's approach of pairing funding with ecosystem support and clear evaluation demonstrates how to de-risk the process for both the city and the startup, while keeping public value at the forefront.

7. Join Communities that Raise the Bar 

You don’t have to do this alone. The GovAI Coalition, for example, now includes roughly 850 public agencies and 2,500 members. The coalition curates policies, vets practices, and shares implementation guides. Borrow what works from others, contribute your own templates, and use the hard-won lessons from your peers in your next pilot.

What “Good” Looks Like in Public Safety Pilots 

When FirstWave Strategies helps agencies and vendors structure pilots for emerging technology, we push for:

• Outcomes that Matter: Focus on metrics like time-to-dispatch, time-to-clear, and officer safety indicators.

• Operator Fit: Train frontline users and supervisors; measure adoption and user-friendliness, not just features.

• Interoperability: Use APIs and data standards that won't lock you into a single vendor later.

• Compliance & Privacy Baked In: Test for disparate impact and document data flows from the start.

• Exit Ramps: Make decisions to stop or scale based on pre-registered metrics.

These practices align with the National Academies' call for careful vetting, measurement, and governance, all while keeping the day-to-day realities of public safety front and center.

A Sample 90-Day Roadmap 

Weeks 0–2: Frame & Guardrails

• Publish 2-3 public safety problem briefs.

• Establish a risk triage and data access plan.

Weeks 3–8: Discovery Sprints

• Invite startups to participate.

• Run short, paid feasibility sprints using sandboxed data and user ride-alongs.

• Capture baseline metrics and success thresholds.

Weeks 9–13: Pilot & Decide

• Run the pilot with frontline training, weekly evaluations, and bias checks.

• Hold a go/no-go session with leadership, legal, and community representatives.

• If the decision is "go," convert to a scale contract with performance and equity clauses.

Bottom Line 

Work with startups the same way you would work the fireground: size up the scene, establish command (governance), deploy in phases, and measure the results. If you're a smaller agency, team up regionally. Cities like San José are proving how a combination of incentives, skills, and governance can speed up learning without spiking risk. The Connective in Phoenix shows that you don’t have to be a mega-city to move together at scale.

Want to learn more? 

FirstWave Strategies helps public safety agencies and vendors design safer, faster paths from idea to impact. Let's talk.