Loading...
Loading...
Updated June 2026
Florida government AI operates at a scale that challenges easy categorization: the third-largest state economy, 22 million residents, 130 million annual visitors, and a disaster exposure profile that makes Florida government the most stress-tested state IT infrastructure in the country after a major hurricane. The Florida Digital Service (FDS), established in 2020 and significantly expanded under DeSantis administration technology initiatives, serves as the state's digital transformation office — focused on modernizing citizen-facing services across agencies that have historically operated on independent legacy systems with limited interoperability. The Florida Department of Economic Opportunity (DEO), since reorganized as the Department of Commerce, became the focal point of Florida's government technology crisis when its antiquated CONNECT unemployment system collapsed under COVID-19 claim volume in 2020 — a failure that cost Florida tens of millions in vendor fees for emergency fixes and became a national cautionary tale for states still running legacy benefits systems. The Department of Business and Professional Regulation (DBPR) licenses 1.3 million professionals and businesses across 30 license types — a citizen-services volume that makes it one of the most automation-ready agencies in state government. The Department of Children and Families (DCF) manages child welfare, substance abuse services, and Florida's public benefits programs under perpetual caseload pressure. LocalAISource connects Florida government clients who need AI partners that understand FDS's technology modernization priorities, the post-CONNECT legacy of Florida benefits system failures, and the seasonal disaster response demands that no other state's government faces at the same frequency and scale.
Florida's CONNECT unemployment system failure in March 2020 is the most important context for any AI vendor approaching DEO or its successor Department of Commerce. The system, built in 2013 on a 1970s-era COBOL foundation, received 100,000+ claims in the first week of pandemic lockdowns — versus its design capacity of 20,000 per day. The system crashed, remained down for days, and ultimately processed claims at a fraction of the required pace for months while the legislature authorized emergency funding for vendor patches. Florida paid approximately $72 million in emergency IT expenditures on CONNECT in 2020 alone, in addition to the $77 million originally spent building it. The reputational damage to DEO, and by extension to Florida's government technology procurement credibility, was severe. The legislative response was the IT modernization funding for CONNECT's replacement (now called the Employ Florida portal, rebuilt on modern cloud architecture) and the creation of the Florida Digital Service with a broader mandate to prevent similar failures across other agencies. FDS's work has been primarily in API development, shared digital infrastructure, and citizen-service modernization — not AI specifically — but its cloud-first architecture standards now govern how agencies deploy technology, which affects AI procurement directly. For DEO's successor functions, the post-CONNECT environment has made fraud detection a political as well as operational priority. Florida paid an estimated $3.5B+ in fraudulent unemployment claims during the pandemic — a number second nationally only to California's $30B+ — and the state has deployed ML-based fraud detection that the Department reports as flagging 9% of new UI applications for enhanced identity verification. The system uses behavioral analytics, cross-agency identity matching (comparing UI application data against DMV, DOR, and voter registration records), and commercial identity verification services. Any vendor approaching Florida UI fraud detection needs to understand that this is a politically visible program — the Governor's office tracks fraud detection metrics quarterly — and that model performance thresholds are set at the executive level, not just the technical level.
The Department of Business and Professional Regulation is one of the most analytically tractable AI targets in Florida state government. DBPR licenses 1.3 million professionals across categories including real estate agents, contractors, medical spas, alcohol beverage distributors, and hotels — a volume that creates clear automation ROI for document review, license renewal processing, and complaint routing. DBPR's 2023 technology modernization project, funded through a combination of state appropriations and licensing fee revenue, included NLP-based application review for contractor and real estate license applications — the two highest-volume categories where missing documentation is the primary source of processing delay. The NLP system classifies incoming application packets, identifies missing required documents before routing to human review, and pre-populates application records with extracted data from uploaded certificates, insurance binders, and examination results. DBPR reports a 28% reduction in applications returned for missing information in the pilot cohort, which translates directly to faster license issuance — a politically sensitive metric in Florida, where contractor licensing backlogs affect construction timelines in a state with a $100B+ annual construction market. DBPR is evaluating AI-assisted complaint investigation routing as its next project: with 80,000+ annual complaints across license categories, routing complexity and duplicate identification are significant staff time sinks. DCF manages child welfare cases for roughly 30,000 children in out-of-home care and administers the ACCESS benefits system (SNAP, Medicaid eligibility, TANF) for 3 million Florida households. DCF's safety decision support tool — built on a predictive modeling framework that Florida has piloted since 2019 — uses ML to stratify child welfare case risk and inform caseworker prioritization. The tool went through multiple equity audits following national concerns about predictive child welfare tools before DCF deployed it in 2022. DCF explicitly frames it as decision support rather than decision automation: caseworkers see risk stratifications as one input, not a determinative output. The ACCESS benefits modernization, funded through CMS Enhanced Match, includes NLP document processing for SNAP renewals and Medicaid eligibility re-determinations — a priority that became acute during the 2023 unwinding process.
No other state in the country has Florida's combination of disaster frequency, population density in vulnerable coastal areas, and institutionalized government disaster response infrastructure. Florida has been under at least one federal disaster declaration in every year since 2016, with major hurricane impacts from Irma (2017), Michael (2018), Ian (2022), and Idalia (2023) producing multi-billion-dollar government response operations. The Florida Division of Emergency Management (FDEM) and Florida Department of Transportation (FDOT) together operate the most sophisticated state disaster response technology infrastructure in the country — and AI is increasingly central to it. FDEM's damage assessment system, rebuilt after Hurricane Irma, uses computer vision on post-storm aerial imagery to classify structure damage at the parcel level — feeding directly into FEMA Individual Assistance eligibility determinations and state-funded disaster relief program targeting. The system processes aerial imagery from fixed-wing and UAS platforms contracted through FDEM's standing disaster response contract, turning raw imagery into structured damage assessments within 48 hours of a storm's passage. Before the AI system, parcel-level damage assessment required 2–3 weeks of field inspector visits — a timeline that delayed relief check delivery when Florida's disaster-affected population most needed rapid support. FDOT's evacuation route optimization AI runs continuously during hurricane season (June 1 – November 30), using real-time traffic sensor data, storm track models from the National Hurricane Center, and historical evacuation traffic patterns to dynamically recommend contraflow implementations and route capacity adjustments. The system's recommendations go to FDOT district engineers for final decision — contraflow on I-75 or I-95 requires human authorization — but the AI model has reduced the time from storm track confirmation to contraflow recommendation from 4 hours to 45 minutes. During Hurricane Ian in 2022, that time savings was operationally significant: the storm's track shifted west faster than forecast models initially showed, and the AI's faster recommendation cycle allowed FDOT to adjust the I-75 contraflow pattern 90 minutes earlier than the prior manual process would have allowed.
Strategic planning for AI adoption, readiness assessment, and roadmap development
Workflow automation using AI, including Make.com-style automation and RPA
Predictive models, data analysis, and ML pipeline development
Text analysis, document automation, sentiment analysis, and language processing
FDS is Florida's digital transformation office, established in 2020 and operating within the Department of Management Services. It manages statewide cloud infrastructure, API standards, and the Florida Civic Technology Marketplace — a catalog of pre-approved technology vendors that agencies can procure from without individual competitive processes. FDS has not issued a standalone AI governance policy, but its API and cloud architecture standards govern how AI systems integrate with state infrastructure. Vendors who want access to the FDS marketplace need to complete a security assessment and demonstrate compliance with Florida's data classification and access control standards. For large AI projects, FDS's technology review team provides a pre-procurement architecture consultation that can identify integration risks before RFP release.
The CONNECT failure produced three lasting changes in Florida IT procurement. First, the state now requires load and stress testing documentation as part of RFP responses for citizen-facing systems — vendors must demonstrate system performance at 5x expected peak load, not just expected load. Second, FDS's architecture standards now prohibit single-vendor locked architectures for critical citizen-service systems, requiring API-first designs that allow component replacement without full system replacement. Third, the Division of Information Technology within DMS now conducts independent technical reviews of major IT procurements before contract award — a step that had been skipped for CONNECT. For AI vendors, these changes mean Florida procurement is slower and more documentation-intensive, but the systems that survive the process are better-scoped and better-supported than pre-2020 deployments.
FDEM's computer vision damage assessment system processes post-storm aerial imagery into parcel-level damage classifications (minor damage, major damage, destroyed) within 48 hours of storm passage — versus 2–3 weeks for field inspector-based assessments. The system feeds directly into FEMA Individual Assistance determinations and state disaster relief targeting. FDOT's evacuation route optimization AI reduces time from storm track confirmation to contraflow recommendation from 4 hours to 45 minutes. FDEM's public shelter capacity modeling tool, deployed after the Ian shelter overcrowding incidents in Collier and Charlotte counties, uses population displacement models and shelter capacity data to recommend pre-storm shelter activation and capacity expansion decisions. Florida has also deployed ML-based FEMA application fraud detection post-Ian, which identified duplicate applications and address anomalies before disbursement.
DBPR's NLP application review system focuses on the two highest-volume license categories: contractor licenses (180,000+ active) and real estate licenses (230,000+ active). The system reduced applications returned for missing information by 28% in the pilot, which translates to faster average licensing — DBPR's statutory processing target is 90 days for contractor applications, and the AI improvement has moved average processing to 62 days for complete applications. For a state where contractor licensing velocity directly affects housing construction timelines, that's a politically and economically significant result. Implementation cost was $165,000 through DBPR's existing IT modernization contract, funded by licensing fee revenue rather than general appropriations.
DCF's safety decision support tool uses ML risk stratification to help caseworkers prioritize which open cases require immediate in-person contact. Before deployment in 2022, DCF commissioned an independent equity audit by a University of South Florida social work research team that tested the model for disparate impact on race, ethnicity, income level, and geographic factors. The audit found no statistically significant disparate impact on race or ethnicity but identified a geographic bias toward lower-resource rural counties where data sparsity caused the model to underestimate risk. DCF addressed this by adding a data augmentation step for counties with fewer than 500 historical cases in the training data. The tool is framed explicitly as caseworker decision support — risk stratification is one input in a human decision process — and caseworkers are trained on both how the tool works and its documented limitations.
Get found by government businesses in Florida.