Government AI Venture Creation

Overview

Tomás Warrick had run the City of Phoenix Office of Digital Innovation for four years when the city manager asked him a question he could not answer: "What has this office actually produced?" Tomás could name the projects - an AI triage tool for the water-complaint hotline, a predictive maintenance model for the fleet, a document-review assistant for the city attorney. But when pressed on which of those were still running, who owned them, and whether any had expanded, the answers were uncomfortable. The triage tool was live but nobody had reviewed its accuracy in eight months. The predictive maintenance model had been paused when its developer left. The document assistant had been used by two attorneys and then forgotten. Phoenix had spent $1.8 million on AI experimentation in three years and had accumulated a graveyard of prototypes. The problem was not bad projects. The problem was no ecosystem.

The Graveyard Problem

Government AI graveyards - collections of pilots that launched and stalled - share a common structure. A problem gets identified. A solution gets funded. A vendor or an internal team builds something. The something works well enough to demo. Leadership celebrates. The team moves to the next problem. Six months later, the something is still technically running, but the person who understood it has left, the data feed it relied on has changed, and the accuracy has drifted without anyone noticing. Citizens are still affected by its outputs, but nobody is checking.

This is not a technology failure. It is a governance failure. Specifically, it is the failure to build the infrastructure that turns individual innovations into durable services. That infrastructure is what an innovation ecosystem is.

The core analogy: a single AI project is a seed. An innovation ecosystem is a greenhouse. The greenhouse does not guarantee that every seed grows. But it provides the consistent conditions - funding, attention, accountability, physical infrastructure - that allow the ones worth growing to survive and produce.

Building the Greenhouse

Tomás rebuilt Phoenix's approach around five structural elements. Each one addresses a specific failure mode he had watched play out.

A dedicated pipeline team

Phoenix's original model assigned innovation work to people who had other primary jobs. The result was that innovation happened when everything else was caught up - which is almost never. The new model created a three-person pipeline team with a single job: move ideas from intake to evaluation to pilot to scaling decision. Three FTEs costs approximately $380,000 per year in loaded salary. Phoenix was spending more than that annually on stalled projects that no one was managing.

Pre-positioned procurement capacity

The most expensive part of government AI innovation is time, not money. A procurement that takes 14 months kills the momentum that makes experimentation possible. Phoenix negotiated two master service agreements - pre-competed contract vehicles - with AI service vendors before identifying specific projects. When an idea cleared evaluation, the team could issue a task order in two to three weeks rather than starting a new solicitation. This does not bypass procurement law; it front-loads the competitive process so that individual experiments can move quickly within an already-vetted structure.

A problem registry

Twice a year, the pipeline team runs structured interviews with frontline staff in each department. Not executive presentations - working sessions with the people who process permits, handle complaints, and manage caseloads. The outputs are entered into a shared problem registry: a database of documented service problems with estimated scale (how many people affected), estimated cost (staff hours, error rates, citizen burden), and feasibility rating (does the agency have the data this would require). When an idea is proposed, it goes into the registry. Ideas already in the registry with strong feasibility ratings move to the front of the evaluation queue.

An explicit scaling gate

Every pilot in Phoenix's ecosystem runs for 90 days and produces a scorecard against pre-specified success criteria. A three-person panel - the innovation director, the department program lead, and the budget officer - reviews the scorecard and makes one of three decisions: scale, extend for another 90 days with a defined test, or stop. The stop decision is documented as a learning outcome, with a summary of what was learned added to the problem registry so the next team that approaches the same problem starts with more information. The fleet maintenance model that had been "paused" was formally stopped, the lessons documented, and the data assets transferred to the team that later built a replacement that actually worked.

Named system ownership

Every AI tool that scales to production gets a named system owner - a department employee, not a vendor contact - whose performance objectives include quarterly reporting on the system's accuracy, usage, and any complaints or errors. The system owner attends the annual review meeting where the innovation panel decides whether to continue, upgrade, or retire each production system. Ownership cannot transfer without a 30-day transition period and a written handoff document. This single change - named owners with actual accountability - is what prevented Phoenix's next generation of tools from joining the graveyard.

Venture Models for City and County Government

The innovation models used by large federal agencies do not scale directly to a city or county with 2,000 to 20,000 employees and a $300 million operating budget. Phoenix is a large city. For smaller jurisdictions, the relevant models look different.

Innovation funds work well at any scale. A competitive internal grant program - $15,000 to $75,000 per award, funded from a budget set-aside - lets department staff propose solutions to their own problems. The grant process is lighter than a full procurement: a two-page proposal, a scoring rubric, a panel review. Successful grantees get both money and one quarter of the pipeline team's time. Failed projects produce a lessons-learned report that goes in the registry. Several Colorado counties have run programs like this with annual budgets between $100,000 and $300,000.

Shared-service platforms are especially powerful for small jurisdictions. A city with 300 employees cannot afford to build its own AI infrastructure. But a consortium of fifteen small cities in the same region can negotiate a shared contract for AI-assisted permit review, or a shared contract for a document-drafting tool, spreading cost across the consortium. The National League of Cities has been actively supporting cooperative procurement agreements for exactly this purpose since 2024.

University partnerships provide technical capacity that small agencies cannot hire. A partnership with a state university's public-policy school or computer-science department can provide graduate student teams to prototype solutions under faculty supervision. The city provides the problem definition and the data access; the university provides the technical execution. The city retains ownership of the resulting system (this must be specified in the partnership agreement). These arrangements typically produce results in four to six months and cost $20,000 to $60,000, mostly in data-access infrastructure and staff time for oversight.

Measuring Ecosystem Health

An innovation ecosystem that cannot report on its own health will not survive the first budget review that asks what it has produced. Phoenix tracks three categories of metrics in a quarterly report to the city manager.

Pipeline health shows the ecosystem is functioning: ideas evaluated, pilots launched, scaling decisions made (with the stop-to-scale ratio), and average time from intake to scaling decision. A healthy pipeline for a city Phoenix's size processes 15 to 25 ideas per year and produces 3 to 6 production deployments.

Service impact shows the ecosystem is producing value: processing time changes, error rate changes, cost per transaction, citizen satisfaction scores - tied to the original problem statement for each scaled tool. Aggregate: "Tools currently in production affect an estimated 34,000 citizen interactions per month."

Equity tracking shows the ecosystem is not automating inequity. For each production tool that affects citizens, Phoenix tracks outcomes disaggregated by neighborhood (as a proxy for income and race) and by language. A tool that processes permit applications 40% faster on average but takes 20 additional days for applications submitted in Spanish has not improved the service - it has made a disparity worse. Equity metrics go into the quarterly report alongside the efficiency metrics.

> The innovation ecosystem's job is not to produce AI. Its job is to produce better public services. AI is one method, not the outcome.

Key Takeaways

• Prototype graveyards are a governance failure. AI tools that launch and stall are not a technology problem. They are the result of missing infrastructure: no named owners, no scaling gates, no accountability for what happens after the demo.
  - Pre-positioned procurement is the time investment that makes experimentation feasible. Master service agreements with pre-vetted vendors allow task orders to move in weeks rather than months. This does not bypass competitive procurement - it front-loads it so individual pilots can move quickly inside a legal framework.
  - A problem registry operationalizes institutional knowledge. Documenting problems with scale, cost, and feasibility ratings - and linking them to lessons learned from prior attempts - means each new team starts with more information than the last. The registry prevents the same problem from being re-discovered and re-failed every three years.
  - Explicit scaling gates prevent both zombie projects and premature cuts. A 90-day evaluation with a documented stop/scale/extend decision treats every outcome as informative. This requires pre-specifying success criteria before the pilot starts, not after.
  - Named ownership is the single most important governance mechanism. Every production AI system needs one named government employee - not a vendor - whose job includes monitoring accuracy, reporting quarterly, and handling complaints. Without ownership, production systems drift toward the graveyard.
  - Small jurisdictions have ecosystem options. Internal innovation funds ($100K–$300K/year), cooperative procurement with neighboring agencies, and university partnerships are all viable models for cities, counties, and small state agencies that cannot afford a dedicated innovation office at scale.
  - Equity metrics belong in every production report. Aggregate performance improvements that mask demographic disparities are not improvements. Track outcomes disaggregated by neighborhood, income proxy, and language from the first pilot forward.