City of Avondale

Local governments are increasingly exploring artificial intelligence to improve service delivery, streamline internal work and help staff make better-informed decisions. AI can support work such as drafting communications, summarizing records, customer service, data analysis and administrative tasks. For cities facing growing expectations and limited resources, it can help create needed capacity.

But innovation and technology alone are not a strategy. Responsible AI adoption requires clear governance, defined accountability, strong data practices and ongoing human oversight. Local government should not use AI simply because it is available. We should use it in ways that are transparent, secure, equitable and aligned with the public interest.

Start with People, Policy and Training

Before making technology decisions, we formed a cross-functional AI Committee. It included representatives from IT, legal, finance, communications, human resources and departments that work closely with data.

Before giving employees access to AI tools, we knew we needed an AI usage policy. Once the policy was approved, we included it in the required employee training. We then rolled out organization-wide training on what AI is, how to apply the policy and how to use both free AI tools and enterprise AI tools responsibly.

We also looked at ROI—or, more realistically, cost avoidance and other benefits. We wanted to understand how an AI Assistant could help employees spend less time on routine work and more time serving customers. In some cases, it may also help departments manage vacancies when appropriate and within budget constraints.

We selected an organization-wide tool that gave all employees access to an AI Assistant. We supported the rollout with communications, training, and full executive leadership support.

This approach helped us put clear structure, guidance and accountability around how we use AI across the city.

Use an AI Usage Policy to Set Clear Guardrails

A practical AI Usage Policy recognizes that generative AI can help with many tasks, but it cannot replace professional judgment. Employees remain responsible for the accuracy, appropriateness and security of the work they submit or act on.

An approved AI environment can:

• Give employees a safe option instead of pushing them toward public consumer tools for work.

• Allow IT to manage licensing, configurations, adoption and user support.

• Create a common platform for policies, training, prompts, templates and use-case guidance.

Training and Change Management

A committee and policy are not enough. Employees need to understand their role in using AI, making decisions with AI support and using the technology correctly.

“Local government should not use AI simply because it is available. It should use AI in ways that are transparent, secure, equitable and aligned with the public interest.”

Clear expectations, defined responsibilities and practical guidance all support the organization’s security posture. Training is not one-and-done. It needs to continue as tools, risks and employee needs change.

Training should address:

• What AI is and what it is not.

• Common risks, including hallucinations, inaccurate summaries, bias, privacy exposure, intellectual-property concerns and prompt injection.

• When employees may use the organization’s approved AI tool.

• What information is safe to enter and what information is prohibited.

• How to review and fact-check AI output before using it.

• When to consult IT, Legal, Records, Procurement, Human Resources or the AI Committee.

• Practical city examples, such as drafting agendas, creating meeting summaries, translating public-facing content, analyzing non-sensitive data and preparing first drafts of communications.

Provide an Approved AI Tool

Providing an approved enterprise AI tool is an important governance decision. It gives employees a practical way to gain productivity benefits while allowing the organization to apply safeguards for data, access, contracts and oversight. We researched and tested a few AI tools before selecting one that fit our budget and gave every employee access, rather than limiting access to a fixed number of licenses.

Start with Low-Risk, High-Value Work

A responsible AI program should show value through controlled, low-risk work before expanding into sensitive workflows.

Examples include:

• Drafting and editing routine internal communications.

• Summarizing public, non-confidential documents.

• Brainstorming public-engagement ideas.

• Creating first drafts of FAQs, procedures or training outlines.

• Assisting with code documentation or test-case development when secure development practices are followed.

• Organizing and analyzing approved, non-sensitive datasets.

AI should not make decisions on its own about eligibility, hiring, discipline, enforcement or services provided to residents. These decisions require staff review and clear safeguards.

Connect AI Governance to Existing City Practices

AI governance should not stand alone. It should connect to the policies and practices governments already use.  We use several cross-functional committees to connect AI governance with existing city practices:

• Data Governance Committee: Sets data strategy and policy, including data classification and related standards.

• Web Governance Committee: Ensures city websites and web content meet accessibility, public-records retention, and security requirements.

• IT Governance Committee: Includes IT and the City Manager’s Office to review proposed and active projects, as well as committee-led initiatives across the city.

These committees intersect because they all rely on data and guide how the city collects, protects, manages, retains and uses it. AI is only as reliable as the information entered into a prompt. Strong data governance, security, records retention and accessibility practices help ensure AI is used responsibly and produces more useful results.

Collaborate Across Government

Government agencies should not treat responsible AI as a competitive advantage developed behind closed doors. We serve many of the same residents, face similar capacity constraints, and share an obligation to protect public trust.

By learning from state partners and peer local governments and by participating in collaborative networks such as the GovAI Coalition, public agencies can build stronger policies, make more informed procurement decisions, and create safer, more useful AI experiences for employees and residents.

Working together helps governments learn from each other instead of each city figuring it out on its own. Shared templates, practical use cases, training resources, and lessons learned can reduce duplicate effort while raising the standard for transparency, security, accessibility, equity, and vendor accountability across the public sector.

Responsible AI in local government is not a one-time policy decision or technology purchase. It is an ongoing practice of testing carefully, training continuously, monitoring outcomes and keeping people accountable for decisions that affect the community.