From the State Government edition of October 29, 2026
EY survey of 200 state and local leaders: 69% have implemented or are implementing generative AI and 96% of adopters report benefits, but 84% worry about shadow AI and only 51% call their governance agency-wide
Ernst & Young LLP (EY US), Center for Government Modernization; fielded by M Booth Research & Insights · State and local government IT modernization and AI adoption (survey) · United States (100 state and 100 local decision-makers)
- Publisher
- EY US
- Original publication
- 2026-10-29
- Source retrieved
- 2026-10-30
What happened
EY's online survey of 200 U.S. state and local decision-makers (100 state, 100 local in jurisdictions of 100,000+), fielded July 23 to Aug. 4, 2026 with a margin of error of plus or minus 7 points, found 69% have fully integrated or are implementing generative AI, and 96% of adopters report a benefit: efficiency (59%), cybersecurity and threat detection (56%) and reduced administrative burden (47%). 84% are concerned about staff using unauthorized or unmonitored tools; 89% worry about dedicated AI funding and 86% about current capability. 94% report AI governance policies or strategies, but only 51% say their approach is comprehensive and agency-wide. 75% expect reliance on legacy systems to cause a significant operational failure, and 38% spend significantly more on legacy maintenance than new solutions. EY's release uses the term 'SI' for AI, noting the survey itself used 'AI'.
Why it matters
Why now: released Oct 29 and not in our archive. It adds a third survey this month, alongside NASCIO and UserEvidence, showing policy on paper running ahead of agency-wide oversight. The state half of the sample is small, and results are self-reported benefits, not measured outcomes.
Evidence and measured results
EY newsroom release opened, including the methodology note; GovTech's report opened for EY's comments on shadow AI and funding. Benefits are respondent perceptions; no measured baselines. State and local results are combined in the release.
Limitations and uncertainty
Commissioned by a consulting firm that sells modernization services; self-reported; online sample of 200 split between state and local; margin of error plus or minus 7 points; full question wording not reviewed.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-29; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- State and local agencies are using generative AI widely but worry about unapproved tools, unstable funding and limited skills, and most lack agency-wide governance.
- Stakeholders
- State CIO and CISO
- Agency chief AI or data officer
- Budget office
- HR and workforce development
- Agency program leaders
- Discovery questions
- Which unapproved AI tools are staff using today?
- Is AI funded as a baseline budget line or one-time appropriation?
- Who enforces the AI policy across all departments?
- Which legacy systems block the AI use cases you want?
- Value hypothesis
- Interpretation: offering an approved, governed tool with training may reduce shadow use and make benefits measurable; the survey reports perceived, not measured, benefits.
- Potential engagement
- Bounded shadow-AI and governance assessment for one agency, ending in an approved-tool rollout plan with measurement and budget justification.
- Unsupported claims
- Benefits are self-reported, not measured
- State-only results not separated in the release
- No evidence that approved tools reduce shadow use
Pre-sales engineering
Role takeaway
- Fit
- Agencies early in governed AI adoption that need to replace unsanctioned tools and connect AI to legacy data.
- Architecture and integration
- Approved enterprise AI tool with SSO, logging and data-loss controls; integration with legacy systems through APIs or data extracts where AI needs program data.
- Prerequisites
- Inventory of current AI tool use
- Data classification rules
- Identity and logging integration
- Legacy system integration assessment
- Constraints
- Unstable dedicated funding
- Legacy systems consuming most IT budget
- Skills gaps
- Combined state and local survey results
- Security
- Detect and limit unapproved AI tools; route staff to approved tools with data controls and logs.
- Proof of value
- Measure shadow tool use before and after an approved tool launch
- Time specific tasks with and without the tool on a sample of staff
- Test whether needed legacy data can be reached securely
Delivery
Role takeaway
- Work
- Move from policy to practice: approved tool rollout, training, compliance checks and benefit measurement tied to budget requests.
- Dependencies
- Dedicated funding
- Trained staff and AI leads
- Legacy system access
- Leadership continuity
- Ownership
- Agency CIO with the state or agency AI governance lead; budget office for sustained funding.
- Skills and adoption
- 86% worry about capability, so training must accompany tool access, with role-specific guidance.
- Governance checkpoints
- Policy applies across all departments, not only some
- Periodic review of unapproved tool use
- Benefit measures reported before scaling
- Acceptance criteria
- Proposed: agency-wide policy coverage confirmed for all departments
- Proposed: measurable drop in unapproved tool use after rollout
- Proposed: documented time savings on at least two defined tasks
- Risks
- Benefits overstated by self-report
- Funding ends after pilots
- Legacy failure disrupts AI-dependent services
- Policy remains unenforced
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Legacy dependency (75% expect a significant failure) means AI projects often sit on top of systems that limit data access, so modernization and AI work compete for the same budget and staff.
Governance
Who approves, reviews and stays accountable for outcomes?
The gap between 94% with policies and 51% with agency-wide approaches mirrors other October surveys; EY's government lead recommends providing approved enterprise tools with agency-wide guardrails to reduce shadow use.
Security and privacy
What data, permissions and controls need testing?
84% concern about shadow AI points to data exposure through unapproved tools; respondents nonetheless report 93% confidence in mitigating privacy and cybersecurity risks, which may be optimistic given the governance gap.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
91% are concerned about attracting and retaining skilled technology talent; 86% about current capability to use advanced AI.
Procurement
What should contracts, pricing and exit terms secure?
89% worry about dedicated AI funding and 88% about modernization funding, often handled as special appropriations; EY's lead urges baseline budgets tied to outcomes.
Operating model
Which teams own the service once it runs?
Only 57% say their modernization roadmap is implemented agency-wide and 51% have comprehensive AI governance, suggesting central policy without consistent departmental adoption.
Publication history
- 2026-10-29State Government · Issue 522 resources
Stable resource ID: ey-sled-modernization-survey-69pct-genai-84pct-shadow-ai-concern-20261029