State and local governments are under increasing pressure to improve outcomes for residents while making every taxpayer dollar count. AI can help, but only if leaders move beyond isolated pilots and focus on rewiring in key service and operations areas to achieve real impact.
Some early AI deployments are already proving successful. Ohio, for example, has cut manual workloads by roughly 50 percent through automated document processing, and New Jersey’s AI assistant led to a 50 percent increase in the number of successfully resolved calls at select call centers.1 Yet adoption remains limited across states nationwide. Only about 6 percent of state chief information officers (CIOs) report having mature, organization-wide AI capabilities.2
The question is not whether AI has potential. It is where and how government CIOs should embed AI to create meaningful results without taking on more complexity than they can manage.
The challenge on the ground
The initial focus for CIOs has been on assembling the tools, platforms, partnerships, procurement structures, and risk guardrails needed to begin experimenting with AI. The next, more difficult challenge is proving measurable business impact and a defensible return on investment by working with agency and business leaders to redesign the processes, behaviors, and workflows that determine how services actually get delivered.
A practical way to do this is to focus on a small number of priority domains: end-to-end areas of public services or operations, such as tax and fee collection, permitting, administration of welfare benefits, facilities maintenance, and workforce training, where leaders can reimagine the full journey across agency lines, rather than automate a single task in isolation. For example, instead of applying AI to verify a single document, a government could redesign the broader assessment-and-certification workflow, improving experience and outcomes across the full process.
The opportunity at hand
Domain-led transformations are twice as likely to achieve impact, making them a stronger path from experimentation to scalable, lasting value. The priority is to choose domains where AI can materially improve performance, the required data and technology capabilities are feasible, and business leaders are prepared to sponsor the necessary process and behavioral changes.
The goal is a “Goldilocks” scope, choosing domains that are ambitious enough to generate meaningful results, but bounded enough to deliver them within six to 12 months. This also creates reusable data and technology capabilities, builds momentum among stakeholders, and supports the change management needed for people to adopt new ways of working.
Three moves can help leaders get there.
1. Redesign the journey before automating it
AI will not fix a fragmented process simply by making one step faster. Government CIOs can partner with business leaders to map the resident or employee journey end to end, use data and stakeholder input to identify bottlenecks, establish shared outcomes and KPIs, and then determine where AI can help simplify, triage, and assist. Permitting redesign in Austin, Texas, offers a useful example. Cross-departmental teams combined shared KPIs with an AI precheck tool as part of an end-to-end redesign, reducing initial review times by 56 percent and nearly halving the review workload.
2. Invest in the workforce, not just the technology
AI adoption is ultimately a change-management challenge, often compounded by a disconnect between business and on-the-ground teams. Frontline staff understand where processes break down, where resident needs are not being met, and where technology may introduce new risks. Businesses and subject-matter experts understand these processes, but that knowledge is often undocumented. Bringing both sides into design early improves the solution and strengthens adoption. Training, new ways of working, and workforce engagement should therefore be treated as core implementation priorities, not activities to add after a tool has been selected.
Data, guardrails, and procurement are where most governments start, and while they remain essential, they do not guarantee impact on their own. The two actions above create the change; the following action helps build the foundations to implement and scale it responsibly.
3. Build the data foundation where it’s needed most
Data readiness is foundational, but it does not have to mean waiting for a multiyear data overhaul. Data remains a major barrier, with 70 percent of high-performing organizations reporting it as the biggest obstacle to realizing AI value. For each priority domain, leaders can identify the few data flows that matter most, establish access and quality requirements, and put governance in place early. This approach avoids a costly pattern in which every agency or program builds its own version of the same data infrastructure. Shared capabilities can make the second and third AI solutions easier to deploy than the first.
Next steps for CIOs
The path to becoming AI-native is not to pursue AI everywhere at once, nor is it to pursue a series of small use cases. Instead, CIOs should apply the Goldilocks discipline and work with business leaders to choose a few high-value domains, define the outcomes and metrics to improve, redesign how work gets done before introducing AI, and then build the enabling capabilities, including data, governance, procurement, and workforce readiness, that make improvement repeatable.
1 “Ohio wins national IT award for AI innovations in unemployment services,” Ohio Tech News, March 6, 2026; “New AI platform empowers state employees to leverage AI to transform government services,” State of New Jersey, July 3, 2024.
2 Harnessing GenAI to Elevate the Citizen Experience, NASCIO and Accenture, September 2025.
This article was originally published on LinkedIn on September 15, 2026.
Hrishika Vuppala is a senior partner in McKinsey’s Southern California office. Jan Shelly Brown is a partner in the New Jersey office. Anant Chouhan is an associate partner in the Chicago office.