Most conversations about AI in government start in the wrong place. They start with the technology — what the models can do, how fast they are improving, what might be possible in five years. For a city manager, an EDC director, or a school district administrator, that conversation is easy to tune out. It sounds like someone else’s problem, or a risk not worth taking on public time.
The more useful conversation starts with the work already sitting on your desk.
At the Elsa Economic Development Corporation, where our practice was first proven in the field, AI is not a pilot program or a line item. It is embedded in the ordinary rhythm of the office: drafting, research, documentation, and analysis. None of it required a technical staff. All of it is reviewable by a human before anything leaves the building. That last point matters, and we will come back to it.
Four places AI earns its keep
Drafting. Public organizations run on written documents: board memos, staff reports, grant narratives, letters of support, public notices. These follow predictable structures, and AI drafts predictable structures well. The practice is not “have the machine write it.” The practice is: give the tool your real context — the project scope, the stakeholders, the timeline, the audience — and let it produce a first draft you then correct. A staff member who once spent three hours on a board memo spends forty-five minutes, most of it on judgment rather than typing. The signature at the bottom is still theirs.
Research and pressure-testing. Before a proposal reaches a council or a board, someone should have asked the hard questions. AI is a patient, tireless first reviewer. Our founder’s standing practice is to reverse the usual arrangement: rather than asking the tool for answers, ask it to interrogate the idea — “Before you respond, ask me twenty questions that would help you understand this project.” The questions surface assumptions, edge cases, and constraints that would otherwise surface in a public meeting. As he has written elsewhere, the tool “doesn’t decide for me. It just makes the boardroom clearer.”
Meeting documentation. Minutes, action items, and follow-up summaries are essential and nobody’s favorite task. Recording a staff meeting (with participants’ knowledge, and consistent with your records policies) and having AI produce a structured summary — decisions made, owners assigned, deadlines set — turns an hour of after-meeting cleanup into ten minutes of review.
Data cleanup and analysis. This is the use case most offices underestimate, and it is where our own story starts. Rivera Educational Consulting grew out of a family educational consulting business with a familiar problem: data scattered across spreadsheets, conditional logic nobody knew how to write, and a master database that existed only in theory. The assumption was that solving it required a programmer. It did not. It required about two hours of asking AI to walk through the problem step by step — then asking it to write plain-language instructions so the rest of the team could use the result without understanding the mechanics. That single exercise saved roughly forty hours of manual work. Business retention surveys, permit logs, sales tax comparisons, enrollment data: most public-sector datasets have a version of this problem, and most of it is solvable the same way.
The discipline that makes it safe
The reason risk-averse organizations can adopt these practices confidently is that none of them remove human judgment. In every case above, AI produces a draft, a question list, a summary, or a formula — and a staff member reviews it before it counts. We teach a simple ordering discipline we call Brain. Train. Execute.: organize your context first, refine the framework with the tool second, and only then scale the output. AI is a multiplier. If an office’s processes are vague, it multiplies vagueness. If they are clear, it multiplies clarity. The preparation is not overhead; it is the safeguard.
There are also places to hold the line. Anything involving personally identifiable information, protected records, or legal determinations deserves explicit policy before any tool touches it. Adoption done well is boring: one workflow at a time, with review built in.
Start with one real task
The gap between organizations using these tools and those waiting for them to “settle down” is widening — not because the technology is exotic, but because fluency compounds. Every drafted memo builds intuition for the next one. Waiting does not pause that process; it means standing still while the baseline moves.
You do not need a strategy retreat to begin. Pick one recurring task — next week’s board memo, last month’s meeting minutes, the spreadsheet everyone avoids — and run it through a tool with proper review. The organizations that feel calmest about AI are not the ones with the most technical staff. They are the ones that started using it for one real thing.
That is the approach we train, and the one we have deployed — including a live program that scored the digital presence of more than 180 businesses for a Texas EDC. Practical, reviewed, and grounded in work that already needed doing.