← All insights

Putting AI on Your Modernization Roadmap: A Practical Guide for State and Local Agencies

Most state and local agencies are not starting from a blank page. They are running decades-old systems, working through modernization projects already in flight, and answering to legislators, auditors, and the public. Now AI has been added to the list.

The temptation is to treat AI as its own initiative with its own budget line. In our experience across government technology, that usually leads to pilots that never connect to anything. A better approach is to put AI where it belongs: on the modernization roadmap you already have.

Start with the systems you actually run

AI is only as useful as the information it can reach. If your case data lives in a mainframe, your constituent records are split across three systems, and key processes still depend on paper forms, those facts shape what AI can realistically do.

Before choosing any AI use, map the basics:

• Which core systems hold the data that matters for your mission?
• Which of those are already scheduled for replacement or upgrade?
• Where is data quality poor, duplicated, or hard to get out?
• Which processes are mostly manual today?

This often shows that the best "AI project" is really a data or integration project that makes later AI use possible. That is not a detour. It is the foundation.

Work with procurement, not around it

Government buying rules exist for good reasons, and they don't move at the speed of product announcements. Planning around that reality saves months.

• Check existing contracts first. AI features may already be included, or addable, in platforms you license today, such as your office suite, CRM, or case management system.
• Bring procurement in early. Ask what vehicles are available to you and how long each path typically takes in your agency.
• Write requirements around outcomes. Describe the problem and the safeguards you need, rather than naming a specific product.
• Ask vendors hard questions. Where is data stored and processed? Is your data used to train their models? What logs and audit trails are available? How do they handle a request to delete data?

Put governance in place before scale

Agencies are accountable for decisions in a way private companies often aren't. Your AI governance should reflect that from day one. It does not need to be heavy, but it should be clear:

• A short agency AI policy covering approved tools, prohibited data, and required human review.
• An inventory of where AI is in use, kept current.
• A simple risk tier: low-risk internal uses (drafting, summarizing) move faster; anything that affects a resident's benefits, eligibility, or rights gets much closer review.
• Named owners for each use, plus a way for staff and the public to raise concerns.

Many states and localities have issued their own AI guidance or executive direction, and the picture keeps changing. Confirm what applies to your agency with your CIO's office and legal counsel before you finalize anything.

A phased roadmap

A roadmap that works in government is usually built in phases, each one earning the trust and funding for the next.

• Phase 1: Foundations. Assess current AI use, publish an interim policy, train staff on safe use, and pick two or three low-risk internal pilots such as drafting correspondence or summarizing public comments for staff review.
• Phase 2: Proven internal value. Expand what worked. Connect AI to cleaner data sources. Build measurement into each use so you can show results in plain terms.
• Phase 3: Resident-facing services. Only after the foundations are solid, consider uses residents see directly, such as improved search or guided help on your website, with clear disclosure and an easy path to a person.
• Ongoing: Review and refresh. Revisit the roadmap on a set schedule. Tools, costs, and rules will change.

Think about funding early

AI work can often be funded as part of modernization rather than as a separate ask. Look at:

• Existing modernization or IT capital budgets where AI supports an approved project.
• Operating budgets for training and policy work, which are usually modest.
• Federal or state grant programs that support technology modernization, data, or workforce training. Eligibility and rules vary, so check with your grants office.

Whatever the source, decision-makers will want to know what problem is being solved, what it will cost to run over time, and how you will know it worked. Build those answers into every phase.