What 5 Public-Sector Engagements Taught Us About AI-Ready Government Data

The federal government published 3,611 AI use cases in 2025, a 105 percent jump in one year, and AI just dethroned cybersecurity at the top of state CIO priorities for the first time in 12 years. Here is what five Thinklytics public-sector engagements (federal, state, county, city, K-12) tell you about which agencies are actually ready, and what the rest need to fix first.

Does the change in administration affect this?

The policy emphasis shifted toward acceleration, but the procurement, disclosure, IP, and oversight requirements (94 government-wide AI requirements identified in GAO-25-107933) are largely intact. The data-readiness work is the same regardless of which administration is naming the priority.

The 2026 NASCIO State CIO survey, published in December 2025, put artificial intelligence at #1 on state CIO priorities for the first time in 12 years. AI displaced cybersecurity, which had owned the top spot since 2014 (NASCIO, December 2025). The federal version of the same shift: the OMB 2025 Federal Agency AI Use Case Inventory documented 3,611 use cases across 56 agencies, a 105 percent year-over-year jump from 1,757 in 2024 (OMB inventory, GitHub repository, March 2026).

Most of that growth is happening at agencies that already had clean transactional data and a governance baseline. The agencies that did not have those have produced the year's most expensive cautionary tales (the SSA AI fraud chatbot reversed after flagging only 2 of 111,000 calls; DOGE-deployed AI hallucinating VA contract values orders of magnitude off). The pattern for 2026 is clear: AI readiness in government is a data-readiness story first.

Here is what five Thinklytics public-sector engagements, one federal, one state, one county, one city, and one K-12, tell you about what AI-ready government data actually looks like.

1. Texas Health and Human Services: $4.2M shifted from AI to data

HHSC was sizing a $12 million AI modernization decision and asked us to assess AI readiness across eight program areas. Past tech projects had failed because data quality issues were ignored at the front of the project. We measured data quality, completeness, process documentation, staff skills, and governance maturity in each program area, and scored them on a fixed rubric.

Three of the eight programs were AI-ready and went straight to deployment scoping. Five needed 6 to 18 months of data and process work before any AI deployment would be safe. We surfaced $7.3 million in automation opportunities across the eight areas. The agency then redirected $4.2 million from AI projects to foundational data work before any deployment. That redirect is the move most public-sector agencies skip and the one most likely to determine whether a 2026 AI program survives an IG review (state-agency-ai-readiness).

The lesson for any state agency reading the NASCIO 2026 results: the AI use cases that move first are the ones in program areas where the data is already clean. Every other program area is a 6-to-18-month data project.

2. Florida Department of Education: 67 districts on one platform in 14 weeks

The Florida Department of Education had 16 weeks to unify student outcome reports from 67 districts for federal IPEDS compliance. Two prior internal attempts had failed over two years because the district student information systems would not align. Our team built a unified data model, seven custom ETL adapters (one per source system), and a validation step. All 67 districts were live by week 14, with zero change orders. IPEDS submission moved from an 8-week cycle to 3 days. Annual reporting labor dropped from $1.9 million to $210,000 (florida-doe-education-analytics).

The data layer that emerged is what makes any future AI use case possible. Predictive enrollment, dropout-risk scoring, equity-gap analytics: each of those requires unified, lineaged district data. The Florida DOE rebuild created the precondition. Most state DOEs are still on the unfunded version of that project.

3. City of San Antonio: $4.1M federal grant funding preserved

A federal audit found data-quality issues across 12 City of San Antonio departments that put $4.1 million in federal grant funding at risk. The city had 90 days. We assessed all 12 departments in the first two weeks, prioritized the top three by funding exposure, built nightly deduplication pipelines for the eight departments with duplicate service records, and stood up a centralized data-quality monitoring system that runs 80 automated checks per night. The city passed the federal re-audit with zero deficiencies and preserved all $4.1 million (city-government-data-quality).

For AI deployment purposes, the relevant artifact is the nightly monitoring system. That is the layer that makes a downstream city AI use case (service-demand forecasting, beneficiary deduplication, real-time grant-utilization reporting) safe to deploy. Cities that try to deploy AI on top of an unmonitored data layer end up reproducing the federal-grant audit failure at a different point in the workflow.

4. Federal Transportation Agency: FOIA from 34 days to 8 days

The agency was processing 2,400 FOIA requests per year, averaging 34 days against a 20-day legal limit, and absorbing $1.2 million per year in penalties. Data was scattered across six regional offices with no central catalog. Staff manually contacted each office for every request. We deployed a centralized data catalog, automated classification of records by FOIA sensitivity and retention, and built a workflow that routed requests directly to the right regional custodians. FOIA response time fell to 8 days within three months. Penalties stopped. The catalog uncovered 18 TB of duplicate data, cutting storage costs by $340K per year (federal-agency-governance).

This engagement is the federal civilian version of "catalog first, AI second." The same catalog that solved FOIA is the catalog that any downstream AI use case at the agency (records review, freedom-of-information triage, document classification) can ride on. Both GAO-25-107933 and GAO-25-107653 effectively recommend this sequencing for federal agencies absorbing the new M-25-22 procurement requirements.

5. Travis County Health Department: 14 manual reports automated, outbreak found 4 days faster

The Travis County epidemiology team was spending 80 percent of their time hand-compiling 14 weekly and monthly Excel reports from four source systems. We consolidated the four sources into a single Power BI dataset refreshed daily, replaced all 14 reports with interactive Power BI dashboards (with drill-downs and automated distribution), and built a disease-surveillance dashboard refreshed every 4 hours during outbreaks. Annual reporting labor dropped from $620K to $80K. Three weeks after launch, during a norovirus outbreak, the team identified the source 4 days faster than the prior weekly cycle would have allowed (county-health-department-analytics).

The 4-day improvement is the single number a county-level executive should keep on a slide. It is also the number that converts the next AI conversation (predictive outbreak forecasting, automated case-investigation prioritization) from a vendor pitch into a budgeted line item. CDC's Public Health Data Strategy 2025-2026 milestones and the $255 million Public Health Infrastructure Grant flowing to STLT agencies (CDC Foundation) make this cohort of work the fastest-funded in the entire public-sector AI ledger right now.

What pattern emerges across all five

Five engagements, one federal, one state, one county, one city, and one K-12 system. Different missions, different regulatory regimes, different funding sources. Every one of them is the same shape.

First, the win is a data layer, not an AI deployment. Florida DOE got 67 districts on one platform. San Antonio got nightly automated checks across 12 departments. The federal transportation agency got a centralized catalog. Travis County got a single dataset refreshed daily. Texas HHSC got an honest assessment of which programs had clean data and which did not. None of those headline outcomes is an AI use case. All of them are the precondition for the next AI use case.

Second, the financial impact is documented in the agency's own measurement system, not in a vendor white paper. $1.9M Florida DOE labor automated, $4.1M San Antonio grant funding preserved, $1.2M federal transportation agency penalties stopped, $620K Travis County labor saved, $4.2M HHSC redirect from AI to data. Each one is in the agency's own audit trail.

Third, the deployment cadence is short. 14 weeks Florida DOE. 12 weeks Travis County. 16 weeks San Antonio. 20 weeks Texas HHSC. 22 weeks federal transportation agency. None is a multi-year program. Public-sector AI readiness work that takes longer than 24 weeks is usually a sign that the front end (inventory + readiness assessment) was skipped.

Fourth, the artifact that lives on after the engagement is operational, not advisory. A working ETL pipeline. A nightly monitoring system. A centralized catalog. A real-time dashboard. A scored readiness assessment with named owners. None of these is a slide deck. The AI use cases that ride on top of these artifacts are the ones that survive an IG review.

What changes for 2026

Three policy and procurement shifts make this pattern more important in 2026 than it was in 2025.

OMB M-25-22 (April 3, 2025) put the new federal AI procurement clock in motion September 30, 2025. Every covered agency must designate a Chief AI Officer within 60 days of the memo and apply the new disclosure, performance, and IP-rights provisions to every AI solicitation (OMB M-25-22 PDF, White House). Agencies that did not document their data layer before the procurement clock started have a harder paper trail to assemble now.

Texas TRAIGA enforcement began January 1, 2026 (Norton Rose Fulbright). The Texas AG can impose up to $200K per violation, and the NIST AI RMF is named as an affirmative defense. Every Texas state and local agency, and every federal contractor handling Texas-resident data, now operates under a regime where data-layer documentation is the prerequisite for the affirmative defense.

NASCIO putting AI at #1 in 2026 means state CIO budgets are following. The state agencies that are AI-ready (clean data, documented lineage, role-based access, governance baseline) move first. The state agencies that are not spend 2026 catching up on the data layer.

The 30-day starting point

For any public-sector executive who has seen the NASCIO 2026 numbers and the federal use-case inventory and is now pressure-tested by their CIO or CAIO to "do something with AI in 2026," the lowest-risk first move is the 30-day AI Readiness Assessment for one program area.

Days 1 to 10: inventory every AI use case (sanctioned and shadow), every dataset they touch, every owner. Days 11 to 20: score the data layer (lineage, quality, access, governance) for the top three use cases. Days 21 to 30: write the AI-readiness brief with a defensible recommendation (deploy, remediate first, or pause) and a 90-day operating plan for the use case selected.

That brief is the artifact that the IG, the state Auditor, the CAIO, and the procurement officer all look at. None of them are looking for an AI strategy slide. They are looking for the documentation that lets the agency defend a deployment.

Frequently asked questions

Does the change in administration affect this?

The policy emphasis shifted toward acceleration, but the procurement, disclosure, IP, and oversight requirements (94 government-wide AI requirements identified in GAO-25-107933) are largely intact. The data-readiness work is the same regardless of which administration is naming the priority.

Can a small county or city skip the inventory phase?

No. The inventory is the cheapest, fastest part of the work and it is the artifact that prevents the SSA-style reversal. A small city should still spend the first week documenting every spreadsheet, dashboard, and shadow GenAI account before scoping anything else.

What's the shortest credible AI deployment timeline for a state agency?

Inventory plus readiness assessment plus one deployed use case end-to-end with full IG documentation is roughly 16 to 20 weeks for a mid-tier state agency. Anyone offering shorter is skipping the documentation that the State Auditor will ask for.

Where do FedRAMP and StateRAMP fit in?

FedRAMP authorization has accelerated for AI cloud services, and Microsoft's December 2025 authorization across the full GenAI portfolio cleared 2.3 million federal employees (Wedbush, December 2025). For sensitive workloads, FedRAMP High is the floor. For state work, StateRAMP is converging.

How does this change for a defense or intelligence agency?

The same pattern applies but the data-classification work is heavier and the procurement vehicle is different (typically EA agreements like the Army-Palantir $10B EA from July 2025 rather than GSA OneGov). The readiness rubric is the same.

If your agency is building its 2026 AI plan and wants the deeper version of this analysis (including the 90-day deployment playbook, the IG-defensible documentation template, and the state-by-state regulatory matrix), our 2026 Government AI Readiness Map is the full operating brief.

Our public-sector practice pages are AI Readiness, Data Foundation, Data Governance Consulting, Analytics & BI, and AI Workflow Automation Consulting. The five engagements above sit at Florida Department of Education, City of San Antonio, Texas Health and Human Services, a federal transportation agency, and Travis County Health Department.

How long does the full set of 5 take in a state agency?

18 to 30 months for a mid-size state agency. Larger federal civilian agencies extend to 24 to 36 months. The pace is set by inter-agency coordination overhead and procurement-vehicle cycle times, not by technical effort.

Which engagement is hardest politically?

Constituent identity resolution. The technical work is well understood; the hard part is reaching agreement across agency leaders on which agency's customer record wins when records conflict. Sponsored escalation from the governor's office or agency head is usually the unblocker.

How does Thinklytics work with public sector clients?

Through prime contractor partnerships on GSA Multiple Award Schedule and 8(a) vehicles. Senior practitioners with experience at state and federal agencies. We do the data foundation work; the prime handles change management and stakeholder rituals.