County Fiscal Resilience
AI opportunities, costs, and safeguards
Continue the fiscal conversation
Use the property-tax scenario you just explored as a discussion starting point. Suppose a rural county is considering two proposals: a developer offers to build a data center, and county staff propose AI tools to improve work productivity.
Your task: What evidence would you require before counting either proposal as part of a financial plan?
Hosting a data center and using AI tools are separate choices within the same fiscal conversation. One does not require the other. Data centers serve multiple types of computing; do not assume every proposal is solely for AI.
Choose your first three evidence requests
Tap up to three priorities. These are the first questions you would ask, not the only safeguards that matter. There is no scored or prescribed answer.
0 of 3 selected. Choose your first evidence requests.
Your discussion brief
Select priorities above to reveal questions and possible evidence providers.
Tell a neighbor: Why these three? What question did you leave out? Who should verify the evidence?
Rural-county lens: What staff capacity, independent advice, utility information, or regional partnership would your county need to assess the proposals?
Connect benefits to the budget
Data-center proposal
Ask for annual county receipts, incentives, county-funded infrastructure, service obligations, and downside scenarios. Distinguish county finances from utility revenues and wider regional benefits.
Investment announced ≠ net county fiscal benefit.
County AI use
Measure a defined task before and during a pilot. Include human review, errors, rework, costs, and service quality. Identify whether saved time releases funds, avoids future costs, or improves service.
Staff time saved ≠ automatic budget savings.
Do not subtract hypothetical benefits from the property-tax scenario gap. Establish comparable scope, timing, and verified assumptions first.
Privacy and security are part of the fiscal decision
Privacy: Is this collection, use, or disclosure of information appropriate and authorized? Security: How is information protected from unauthorized access, alteration, or loss?
Try the resident-complaint example
A staff member wants AI to summarize complaints containing names, addresses, and personal circumstances. Before any upload, identify the necessary fields, approved environment, authorized purpose, vendor terms, access rules, and reviewer. Use synthetic examples for this workshop.
Ask how the proposed safeguards affect costs, workload, service quality, and public trust. An apparently cheaper tool may require more review or controls.
Design a small UF/IFAS–county pilot
Discuss one task, one benefit to measure, one safeguard, and one accountable county owner. Potential Extension contributions include evidence interpretation, staff education, pilot evaluation, and resident communication. Technical or legal determinations need the appropriate county and specialist reviewers.
Closing prompt: “Before expanding this pilot, we would need to demonstrate ______ while protecting ______.”
Full evidence menu and background sources
- Net county revenue and timing: What reaches the county budget after incentives and county costs, in which years, and under which tax assumptions?
- Infrastructure and who pays: What power, cooling-water, wastewater, road, and emergency-service requirements apply to this design? Who funds upgrades and ongoing obligations?
- Jobs and community benefits: How many construction versus ongoing jobs, with what skills and wages? What local procurement and neighborhood effects are supported by evidence?
- Enforceable commitments and downside: What happens if construction is delayed, operations shrink, or the facility closes? Which commitments, remedies, and monitoring arrangements are enforceable?
- Measured productivity and service quality: For a defined task, does a pilot improve completion time and accuracy after review and rework? How would saved time improve services or change actual spending?
- Full AI adoption cost: What are the costs of licenses, integration, training, review, security, maintenance, and exit? Which benefits remain after those costs?
- Privacy and appropriate data use: Which data are necessary and authorized? Can identifiers be removed? What do vendor terms say about retention, reuse, training, deletion, and subcontractors?
- Security, oversight, and accountability: Who can access the system? How are incidents handled? Who verifies outputs, approves consequential decisions, monitors performance, and can stop use?
These questions are workshop prompts, not claims that a proposal will produce a particular benefit or harm. No new statewide dataset, tax estimate, legal analysis, or fiscal calculation is added.
NIST AI Risk Management Framework
Voluntary risk-management background; not a county legal determination. Page checked October 22, 2026.
NIST Generative AI Profile (2024)
Background on generative-AI risks and risk-management actions.
DOE: 2024 data-center energy report announcement
Historical national energy context, published December 20, 2024; not an estimate for a Florida facility. Page checked October 22, 2026.