State and local governments sit at the front line of the citizen experience, delivering everything from unemployment benefits and driver licensing to permits, tax administration, public safety, and emergency response. They are the level of government that most citizens actually touch, far more often than any federal agency, which makes the quality of their service delivery a defining factor in how the public perceives government as a whole. Yet these agencies are being squeezed from every side by rising citizen expectations shaped by private sector digital experiences, persistent workforce shortages, tight fiscal envelopes, decades of accumulated technical debt in aging legacy systems, and increasingly complex operational environments Public trust adds urgency to the challenge, with the Edelman Trust Barometer showing global trust in government moving only from 44% in 2016 to 53% [1] in 2026, keeping transparency, accountability, and data integrity under a persistent spotlight.
Against this backdrop, artificial intelligence has moved from being a talking point to a budgeted priority, reframing how governments design, deliver, and measure public services. The shift is already visible in the numbers. NASCIO reports that 88% of states have rolled out AI responsible use policies, and that Gen AI adoption among state CIO staff climbed from 53% in 2024 to 82% [2] in 2025. Local and state agencies show the same direction of travel, though from a lower base. EY’s 2025 state and local government survey identifies AI as the leading area of technology adoption, with 45% of IT decision-makers reporting its implementation and 39% [3] deploying Gen AI Citizen services, infrastructure management, and compliance automation account for much of that activity. The strategic question is no longer whether governments should adopt AI, but how quickly and how responsibly they can move from isolated pilots to trusted, operational AI that delivers measurable public value.
The pivot to AI is not a single procurement decision; it is a structural response to converging pressures that legacy operating models can no longer absorb. Three forces, in particular, are driving the shift.
- Workforce shortages and fiscal constraints: Public agencies continue to face persistent hiring gaps and flat budgets, with recovery in the government labor market lagging in most other sectors. An aging workforce is compounding the problem as experienced staff retire, and institutional knowledge walks out the door, often faster than it can be replaced. With head count constrained and budgets under scrutiny, agencies are turning to intelligent automation and AI assistants to absorb repetitive, high-volume work so that scarce employees can focus on judgment-intensive, citizen-facing, and policy tasks that genuinely require human expertise.
- Rising, experience-driven citizen expectations: Residents conditioned by seamless commercial digital services increasingly view bureaucratic friction, such as rigid business hours, opaque processes, long wait times, and fragmented information spread across disconnected departments, as institutional failure. They expect to interact with government in the way they interact with a bank or a retailer: instantly, in their own language, and on their own schedule. Resident-facing tools such as chatbots, virtual assistants, and streamlined service interfaces are consistently ranked among the highest-potential AI applications by local practitioners because they precisely close this expectation gap.
- Generative and agentic AI reaching production maturity: Gen AI has crossed from experimentation into daily use, and agentic AI, capable of planning and executing multistep tasks with minimal human input, is now entering production. NASCIO notes that eight states already report agentic AI tools in production, signaling a clear next phase beyond content generation. This maturation matters because it transforms AI from a drafting aid into an operational actor capable of completing workflows end-to-end, provided the right governance guardrails are in place.
For most of the past three years, AI in government meant a pilot: a single chatbot, a proof of concept, or a working group producing recommendations. In 2026, that is changing as agencies move the more reliable tools into daily operations, ending what practitioners now call “pilot purgatory,” where promising experiments stall indefinitely without ever reaching production. This transition marks a decisive move beyond AI experimentation toward trusted, operational AI that enhances citizen services, workforce productivity, and decision-making, with pension administration emerging as an early example of AI scaled successfully into core operations. The Government AI Landscape Assessment 2026 describes this progression as a four-stage journey that most states are still in the early stages of navigating [4].
| 1. Readiness | 2. Piloting | 3. Implementation | 4. Impact |
| Governance frameworks, responsible use policies, data foundations, and workforce readiness are established. | Gen AI proofs of concept test what works and what does not before operational rollout. | Reliable tools move into daily operations, delivering measurable outcomes for citizens and productivity. | Outcomes are measured and fed back, strengthening governance and fueling the next cycle of adoption. |
The maturity arc is consistent across jurisdictions, but progress is uneven. Most states are still building governance and piloting Gen AI; a smaller group is beginning to scale; and very few have fully embedded AI into core operations with robust, continuous measurement. Crucially, the journey is iterative rather than strictly linear. A state that has scaled Gen AI may still be at the readiness stage for agentic AI, needing to revisit governance, data, and skills before it can automate more complex workflows. The differentiator is, therefore, no longer access to models, which are now widely available and increasingly commoditized; it is the institutional discipline to operationalize them responsibly, measure their impact, and improve continuously.
AI’s impact is concentrated in five domains that mirror the demand-side priorities identified in Avasant’s State and Local Government Digital Services 2026 Market Insights™: citizen engagement, benefits and claims, security operations, legacy and platform modernization, and workforce augmentation. Together, they span the full arc of public service, from the citizens’ first interaction to the back-office systems and people that fulfill it.
Resident-facing assistants are the most visible and fastest-scaling use case. Nebraska’s Resident AI Assistant, launched at the Department of Motor Vehicles in January 2026, drew more than 38,000 users (59% of DMV website visitors) who asked over 88,000 questions in more than 50 languages, with nearly 19% of interactions occurring outside normal business hours, precisely when live staff is unavailable. The pilot cut driver and license services call volume by up to 20% [5], made more than 3,400 referrals to online services, and is now being deployed statewide across additional agencies. California is running parallel Gen AI pilots, including a multilingual constituent-communication translator that has proven especially valuable in a linguistically diverse state where the cost of professional human translation had been the binding constraint on equitable access to services. The common thread is that AI extends the reach of government, offering more hours, more languages, and more channels, without a proportional increase in head count.
High-volume, rules-bound workflows are prime candidates for Gen AI because they combine repetitive effort with clear policy logic and significant consequences for citizens when they go wrong. North Carolina’s Division of Employment Security launched one of the state’s first public-facing Gen AI solutions, an intelligent virtual assistant that helps residents navigate the unemployment insurance claim process, built on AWS after an eight-year cloud journey that now sees more than 80% [6] of workloads in the cloud. Test deployments in New Hampshire and Colorado similarly use Gen AI to help adjudicators settle claims faster while improving accuracy, demonstrating how prediction and drafting can compress cycle times without removing human oversight. For services that families depend on during moments of crisis, such as job loss, disability, or housing instability, faster and more accurate processing is not merely an efficiency gain; it is a direct improvement in citizen welfare.
As agencies digitize, the attack surface widens, and the public sector remains a favored target because of the sensitive resident data it holds. The 2025 IBM Cost of a Data Breach Report puts the average public sector breach at $2.86 million, up roughly 12% [7] year over year, with identity compromise and citizen data exposure among the most common consequences. The same report finds that public sector organizations take an average of 202 days simply to identify a breach, well above the global average, giving attackers a long window to inflict damage. Governments are responding by embedding AI-enabled security operations, zero-trust architectures, and automated threat detection into core operations, using AI defensively to detect anomalies and contain incidents faster. This is elevating cyber resilience from an IT concern to a board-level and, increasingly, a public-trust priority, since a single high-profile breach can undo years of goodwill in digital services.
Operational AI depends on a modern data foundation, and this is where many governments face their hardest work. State and local IT spending in the US is projected to reach $160.2 billion in 2026, a 4%–6% [8] increase over the prior year, yet much of that investment still sits on top of centralized, siloed, reconciliation-heavy systems, with critical records locked inside decades-old mainframe applications. Because AI is only as reliable as the data beneath it, agencies are rebuilding legacy estates around cloud-native, interoperable platforms that support secure data sharing, digital identity, and integrated service delivery across departments and jurisdictions. That connective layer is what allows AI to move past a single agency into genuinely connected government, where what a citizen shares with one department carries over to the next instead of being collected again at every door.
AI is reshaping how public employees work rather than simply replacing them. A 2026 MissionSquare and PSHRA workforce survey found that more than half of state and local workers reported AI had improved their productivity and the quality of their work, while public sector HR teams are already using AI to draft interview questions (45%), write job descriptions (42%), and drive process improvement (30%) [9]. As routine drafting, summarizing, and data entry tasks are automated, the role of the public servant is shifting from manual processing toward exception handling, decision validation, and higher-touch citizen engagement. This has an important secondary benefit: by removing the most tedious parts of public sector jobs, AI can make government roles more attractive and help agencies retain talent in a tight labor market, provided investment in reskilling keeps pace with the technology.
A range of live deployments illustrates how AI is delivering measurable outcomes across jurisdictions and service lines, from resident assistants to agentic automation.
| Jurisdiction | Application | Measured outcome | Enabling partner |
| Nebraska [5] | DMV “Ask the DMV” Resident AI Assistant; 24/7 multilingual citizen support | Over 38,000 users, about 88,000 questions in more than 50 languages; up to 20% call-volume reduction; now going statewide | Tyler Technologies |
| North Carolina [6] | Gen AI virtual assistant for unemployment insurance claims navigation | One of the state’s first public-facing Gen AI tools; over 80% of DES workloads now in the cloud | AWS (Amazon Bedrock) |
| California [10] | Concurrent Gen AI pilots: multilingual constituent translation, DMV plain-language explanations, Caltrans incident triage | Translation pilot expanding equitable access; human-in-the-loop review across use cases | Multiple/state-run |
| New Hampshire and Colorado [11] | Gen AI to speed unemployment claim adjudication and boost adjudicator productivity | Faster settlement and improved accuracy in claims processing | GovTech/state programs |
| Multiple US States (eight jurisdictions) [2] | Agentic AI tools automating complex, multistep workflows | Already reported in production, signaling the move beyond generative AI | NASCIO-tracked |
The same connectivity that enables faster, smarter service delivery also widens exposure to data quality failures, model drift, and biased or unexplainable decisions. These risks carry particular weight in government, where decisions affect people’s benefits, liberty, and livelihoods, and where the public has a right to understand how those decisions are made. NASCIO’s 2025 data point to growing state-level action: 88% of states have AI responsible use policies, 84% are inventorying AI uses across agencies, and 82% [2] have created advisory committees or task forces to steer adoption. Even so, meaningful gaps remain. Only a minority of agencies currently test AI applications for bias or train all staff on responsible use, leaving room for inconsistent or unaccountable deployments. Public records law, procurement rules, and strict limits on where sensitive resident data may reside shape every deployment decision. It is no coincidence that the pilots that succeed consistently share three traits: a clearly measurable metric, a defined human-in-the-loop checkpoint, and a published procurement playbook that others can learn from and replicate.
AI should be treated as a long-term investment in public service capability, not as a series of disconnected experiments chasing the latest tool. As governments modernize systems and adopt sovereign cloud, connected data ecosystems, and AI, the focus must shift decisively toward measurable outcomes. Success will depend less on the sophistication of any single model than on cross-agency coordination, disciplined governance, workforce enablement, and sustained attention to citizen value. Four imperatives will separate leaders from laggards over the next 24 months.
- Prioritize high-volume, rules-bound workflows first: Concentrate early deployments on citizen inquiries, benefits navigation, permitting, and document processing, where volumes are high, rules are clear, and time savings are quickly measurable. These quick, defensible wins build the evidence base and political capital needed to justify broader, more ambitious adoption.
- Embed AI within enterprise modernization, not alongside it: Align AI initiatives with cloud migration, data governance, and digital identity rather than running them as isolated innovation projects. Standalone pilots dilute value and create integration debt; AI scales only on a modern, interoperable data foundation that treats data as a shared enterprise asset.
- Establish governance before scale: Define responsible AI policies, bias testing, human-in-the-loop checkpoints, transparency standards, and data residency rules upfront rather than retrofitting them after an incident. Governance maturity, not model access, is the true constraint on scaling AI in the public sector and the surest safeguard of public trust.
- Measure impact and enable the workforce: Set baseline KPIs such as cycle time reduction, call deflection rates, cost savings, and citizen satisfaction scores before deployment so value can be proven and refined. In parallel, invest in reskilling so employees move confidently from processing to oversight, exception handling, and engagement, turning the workforce into a partner in transformation rather than a casualty of it.
The early evidence from Nebraska, North Carolina, California, and states already running agentic AI in production shows that operational AI can deliver measurable value. Yet progress remains uneven, with many agencies constrained by legacy systems, limited budgets, and governance frameworks still taking shape. The governments that succeed will not be those that adopt AI fastest or pursue every new capability, but those that scale it responsibly through modern data foundations, disciplined governance, workforce readiness, and clear measures of citizen impact. Done well, AI can improve the efficiency and resilience of public services while strengthening the trust on which effective government depends.
[1] https://www.edelman.com/trust/data-dashboard
[2] https://www.nascio.org/wp-content/uploads/2026/03/NASCIO_Agentic-AI-Report_2026_.a11y.pdf
[8] https://www.govtech.com/budget-finance/what-will-state-and-local-government-spend-on-it-in-2026
[9] https://pshra.org/2026-state-and-local-government-workforce-survey-putting-ai-to-work-in-hr/
[10] https://precisionaiacademy.com/insights/state-local-government-ai-pilots-2026
By Devansh Vyas, Research Analyst, and Eratha Poongkuntran, Associate Director, Avasant
Avasant’s research and other publications are based on information from the best available sources and Avasant’s independent assessment and analysis at the time of publication. Avasant takes no responsibility and assumes no liability for any error/omission or the accuracy of information contained in its research publications. Avasant does not endorse any provider, product or service described in its RadarView™ publications or any other research publications that it makes available to its users, and does not advise users to select only those providers recognized in these publications. Avasant disclaims all warranties, expressed or implied, including any warranties of merchantability or fitness for a particular purpose. None of the graphics, descriptions, research, excerpts, samples or any other content provided in the report(s) or any of its research publications may be reprinted, reproduced, redistributed or used for any external commercial purpose without prior permission from Avasant, LLC. All rights are reserved by Avasant, LLC.
Login to get free content each month and build your personal library at Avasant.com