{"resourceId":"nz-cross-agency-ai-survey-2026","versions":[{"version":"external-7d164862411ee663d8ba1aeb48feec09c9e096dca0c4c1c174b588df05e25bc3","resource":{"id":"nz-cross-agency-ai-survey-2026","title":"New Zealand survey shows operational expansion while effectiveness remains self-reported","organization":"New Zealand Government Digital Delivery Agency","sector":"Government shared services and workforce","geography":"New Zealand; transferable governance lessons for U.S. states","publishedAt":"Last updated August 19, 2026","publicationDate":null,"eventDate":null,"sourceName":"2026 cross-agency survey of use cases for artificial intelligence","sourceLabel":"Government survey report","sourceUrl":"https://www.digital.govt.nz/dmsdocument/264~report-2026-cross-agency-survey-for-artificial-intelligence-ai-use-cases/html","evidenceClass":"government-evaluation","outcomeClass":"emerging","topics":["knowledge-work","developers-agents","infrastructure","data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"Agency reporting indicates more operational AI use, but adoption counts do not establish causal service benefits.","sledRelevance":"A comparator for state shared-service portfolios, not a U.S. state census or a transferable legal framework.","evidence":"59 organisations reported 545 use cases, including 167 operational cases. The prior survey had 70 organisations and 272 cases. Benefits are agency-reported; changing respondents complicate trend interpretation.","architectureImplications":"Interpretation: record lifecycle stage, dependencies and model permissions in a common portfolio record; do not infer an architecture from adoption counts.","governanceImplications":"Interpretation: validate portfolio entries against actual workflows and give every claimed benefit an evidence owner.","securityPrivacyImplications":"The report identifies reliability and data sovereignty as remaining challenges. Interpretation: verify boundaries locally rather than treating reduced concern as reduced exposure.","caveats":"No controlled baseline, measured time savings, response-rate denominator or common outcome rubric supplied. August update is not a September measurement. Jurisdiction and survey composition limit transfer.","streamIds":["state-government"],"roles":{"sales":"Interpretation: Discuss portfolio visibility with the state CIO, agency leaders, workforce representatives and budget owners. Ask how many listed uses actually operate, who benefits and what evidence justifies continuation. A bounded inventory validation and measurement engagement could help distinguish scaling candidates from unsupported experiments. The value hypothesis is better allocation of support and implementation effort. Do not translate overseas adoption growth into a customer's likely savings or maturity ranking. Identify which agencies can supply comparable records and whether a shared service would meet their differing responsibilities, access needs and funding arrangements.","engineering":"Interpretation: Fit is portfolio instrumentation across copilots, developer tools and bounded agents, rather than a particular model deployment. Link each use case to its environment, data class, connectors, human authority, version and telemetry. Prerequisites include agreed lifecycle definitions and access to actual deployment records. Validate a sample against system configuration, test tenant boundaries, and rehearse service fallback. A proposed proof of value should compare one workflow's quality, effort and operating cost with its existing baseline. National survey findings provide no capacity specification; local latency, sovereignty, security and legacy-integration requirements remain decisive.","delivery":"Interpretation: Start by reconciling agency inventories with operational owners and selecting a small evaluation cohort. A central portfolio lead maintains common definitions; agency service owners remain accountable for outcomes. Dependencies include analyst capacity, staff participation and approved handling of evaluation records. Train contributors to distinguish usage, sentiment and observed benefit. Proposed acceptance criteria: each sampled operational use has a named owner, verified lifecycle status, baseline and scheduled continuation decision; unresolved sovereignty or accessibility failures have an approved disposition. Risks include respondent selection, inconsistent definitions and continued funding based on activity rather than public value."},"retrievedAt":"2026-09-09T03:01:07Z","enrichedAt":"2026-09-09T03:01:41Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: test usefulness by role and accessibility needs, including staff who do not adopt the tool.","procurementImplications":"Interpretation: buy reusable capabilities only after checking program fit and lifecycle support obligations.","operatingModelImplications":"Interpretation: portfolio staff should separate planned, trial and operational uses and schedule evidence reviews.","updateExplanation":"No matching survey URL in the complete archive or targeted cross-agency search. August material newly added for the changing survey denominator and operational-stage distinction; no fresh September outcome asserted.","sourceVerification":{"openedUrl":"https://www.digital.govt.nz/dmsdocument/264~report-2026-cross-agency-survey-for-artificial-intelligence-ai-use-cases/html","referenceExcerpt":"More complex challenges remain, particularly around public acceptance of government AI use, AI reliability and data sovereignty.","promptVersion":"sled-research-v3.1","model":null,"basis":"agent-reported inspection"}}}]}