AI Transformation opportunity
A portfolio of AI-enabled projects sized against Arkansas's actual IT, procurement, and operations baselines — designed to automate state operations, catch fraud, waste, and abuse, find contract savings, and replace expensive legacy software. This is additive to the existing $300M / 6-year Arkansas Forward target and builds directly on the AI Center of Excellence's February 2026 framework. Every project is structured as a copilot or decision-support pattern, consistent with Act 848 of 2025's human-in-the-loop requirement.
Four-pillar strategy
Fraud, Waste & Abuse
Stop improper payments before they go out the door — Medicaid, SNAP, UI, tax, procurement.
Contract & Spend Savings
Use AI to inventory, benchmark, and renegotiate the $830M IT and $1.7B procurement base.
Operations Automation
Replace high-volume manual workflows — FOIA, permits, citizen service, tax filing, HR.
Legacy Software Replacement
Modernize the largest legacy contracts (ARIES, MMIS, ARCSIS) on the federal modular pathway with AI-accelerated migration.
Predictive Analytics
Move from reactive to predictive — cybersecurity, transportation maintenance, building energy.
Project portfolio
| Project | Agency | Est. annual savings ↓ | Timeline | Complexity | Confidence | |
|---|---|---|---|---|---|---|
Medicaid Claims FWA AI sprintFraud, Waste & Abuse | DHS / Division of Medical Services + Medicaid Inspector General | $30.0M–$60.0M per year | 12-24 months | Medium | High | |
AI-Assisted Procurement Spend Analysis & Contract Optimization sprintContract & Spend Savings | DFA Office of State Procurement + SAS / OST | $25.0M–$60.0M per year | 9-15 months | Medium | High | |
MMIS Modular Replacement Legacy Software Replacement | DHS / Division of Medical Services | $20.0M–$40.0M per year | 24+ months | High | Moderate | |
Tax Fraud & Underpayment Detection Fraud, Waste & Abuse | DFA / Office of Income Tax + Office of Excise Tax | $15.0M–$40.0M per year | 12-24 months | High | Moderate-High | |
Software License Optimization AI Contract & Spend Savings | SAS / Office of State Technology Services (OSTS) | $15.0M–$40.0M per year | 3-6 months | Low | High | |
Enterprise Generative AI Workforce Assistant sprintOperations Automation | All cabinet (anchor: SAS / Office of State Technology) | $15.0M–$35.0M per year | 3-6 months | Low | Very high | |
ARIES Modular Modernization Legacy Software Replacement | DHS / Division of County Operations | $15.0M–$30.0M per year | 24+ months | High | Moderate | |
SNAP Eligibility Error Reduction sprintFraud, Waste & Abuse | DHS / Division of County Operations | $10.0M–$25.0M per year | 12-24 months | Medium | High | |
AI Citizen Service Agents (Voice + Chat) Operations Automation | DHS, DFA, DMV (cross-cabinet) | $10.0M–$25.0M per year | 6-12 months | Medium | Moderate | |
ARDOT Predictive Maintenance Predictive Analytics | Arkansas Department of Transportation | $10.0M–$25.0M per year | 12-24 months | High | Moderate | |
Cooperative Purchasing Optimizer Contract & Spend Savings | SAS / OSP | $8.0M–$20.0M per year | 6-12 months | Medium | Moderate-High | |
Caseworker Decision Support (DHS, DOC, DWS) Operations Automation | DHS + Department of Corrections + Division of Workforce Services | $8.0M–$20.0M per year | 12-24 months | Medium | Moderate-High | |
Procurement & Vendor Fraud AI Fraud, Waste & Abuse | SAS / Office of State Procurement | $5.0M–$20.0M per year | 9-15 months | Medium | Moderate-High | |
Unemployment Insurance Fraud Detection (scale-out) sprintFraud, Waste & Abuse | Department of Commerce / Division of Workforce Services | $5.0M–$15.0M per year | 6-12 months | Medium | Very high | |
Duplicate Contract Detection Contract & Spend Savings | SAS / OSP | $5.0M–$15.0M per year | 3-6 months | Low | Moderate-High | |
AI Cybersecurity SOC Predictive Analytics | SAS / Centralized Cybersecurity Office (per EO 25-10) | $5.0M–$10.0M per year | 12-24 months | High | Moderate-High | |
AI Permit Triage & Routing Operations Automation | All cabinet permitting agencies | $3.0M–$8.0M per year | 6-12 months | Medium | High | |
AI-Assisted Tax Return Processing Operations Automation | DFA / Revenue Services Division | $3.0M–$8.0M per year | 9-15 months | Medium | Moderate-High | |
ARCSIS (Child Support) Modernization Legacy Software Replacement | DFA / Office of Child Support Enforcement | $3.0M–$8.0M per year | 12-24 months | High | Moderate | |
State-Buildings Energy AI Predictive Analytics | SAS / Building Authority | $3.0M–$8.0M per year | 6-12 months | Medium | Moderate-High | |
Workers Comp Fraud AI Fraud, Waste & Abuse | Workers Compensation Commission / Insurance Department | $2.0M–$8.0M per year | 6-12 months | Low | Moderate | |
AI Contract Redlining for ARBuy Contract & Spend Savings | SAS / OSP | $2.0M–$6.0M per year | 3-6 months | Low | High | |
Fleet Telematics & AI Optimization sprintPredictive Analytics | DFA / Office of Vehicle Services (cross-agency) | $3.0M–$5.0M per year | 9-15 months | Medium | High | |
AI FOIA Response Hub Operations Automation | SAS — cross-cabinet | $2.0M–$5.0M per year | 6-12 months | Medium | Moderate-High | |
AI HR / Onboarding Automation Operations Automation | SAS / Office of Personnel Management | $2.0M–$5.0M per year | 3-6 months | Low | Moderate-High | |
State P-Card & Travel Anomaly AI Fraud, Waste & Abuse | SAS / Office of Accounting | $1.0M–$3.0M per year | 3-6 months | Low | Moderate-High | |
Document Remediation AI (Section 508) Operations Automation | All agencies (web + publications) | $1.0M–$3.0M per year | 3-6 months | Low | High | |
AI Translation / Multilingual Services Operations Automation | All citizen-facing agencies | $1.0M–$3.0M per year | 3-6 months | Low | High |
Implementation roadmap
Scoping sprint — 6 priority projects
Six go/no-go feasibility memos for the top six projects (the sprinttags in the table). Each memo: a quantified target savings range against Arkansas-specific data, vendor-versus-build comparison, data readiness assessment against the Arkansas Data Hub, Act 848 compliance review, implementation cost estimate, and a recommended go/no-go.
~$400K-$700K external advisory + 1.0-1.5 FTE; no new procurement required to begin.
Narrow pilots in parallel
Each project that survives Phase 1 enters a 3-6 month pilot, deliberately scoped narrow: a single Medicaid claims category, 2-3 SNAP county offices, a single tax type, a 200-300 employee workforce-assistant cohort. Narrow pilots produce measurable savings within the fiscal year and surface implementation issues before scale-out commits significant capital.
Expect 1-2 pilots to wash out — that's by design, not failure.
Production scale-out
Anchored by the Workforce Assistant rollout, which touches every department and creates the user-facing familiarity that makes the more technical projects easier to land. Pennsylvania scaled its ChatGPT Enterprise pilot from 175 to 3,000+ users in 12 months. Longer-cycle projects (citizen portal, legacy modernization, real estate) sequence behind the anchor.
Full portfolio decision by month nine; every project in pilot, production, or formally retired by end of year two.
Governance & accountability
Governance sits inside the existing Arkansas Forward project management organization rather than being stood up separately. The AI & Analytics Center of Excellence remains the technical authority and approves individual pilots against the NIST AI Risk Management Framework. The AI CoE-recommended Chief Artificial Intelligence Officer role should be filled before or alongside the Phase 1 sprint.
Each project reports monthly against three metrics: dollars saved or recovered, percent of milestone completion, and the AI CoE's safety and bias review status. Quarterly reporting aggregates into a single Arkansas Forward AI dashboard that mirrors the existing Arkansas Forward Master Initiative Tracker.
Risks & mitigations
Vendor capture
Any of the larger projects can fall into vendor-dependent multi-million-dollar commitments before pilot results are in. Mitigation: structured Phase 1 vendor-versus-build comparison and time-limited pilot scope.
Act 848 compliance
Every consequential AI decision must keep a human in the loop. The portfolio is structured around copilot and decision-support patterns rather than autonomous decision-making, which keeps every project within the statute.
Data readiness
Several projects depend on the Arkansas Data Hub being production-ready for sensitive data flows. The Phase 1 sprint validates this dependency before any project commits.
Public messaging
AI fraud detection in Medicaid and SNAP is politically visible. Communicate as program integrity that protects benefits for the people who actually need them.
Hiring competitiveness
A Chief AI Officer and a small technical team must be hired or contracted. The current state pay plan may not be competitive for these roles without a targeted IT pay-band exception.