{"resourceId":"pennsylvania-chatgpt-pilot","versions":[{"version":"legacy/2026-08-29/pennsylvania-chatgpt-pilot","resource":{"id":"pennsylvania-chatgpt-pilot","title":"State workforce pilot reports large perceived savings but uneven readiness","organization":"Commonwealth of Pennsylvania","sector":"State government","geography":"Pennsylvania, United States","publishedAt":"March 2025","sourceName":"Lessons from Pennsylvania's Generative AI Pilot with ChatGPT","sourceLabel":"Commonwealth pilot report","sourceUrl":"https://www.pa.gov/content/dam/copapwp-pagov/en/oa/documents/programs/information-technology/documents/openai-pilot-report-2025.pdf","evidenceClass":"government-evaluation","outcomeClass":"mixed","topics":["knowledge-work","data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"Pennsylvania equipped 175 employees across 14 agencies with ChatGPT Enterprise for a yearlong pilot using surveys, focus groups, office hours, and role-specific support.","sledRelevance":"The pilot offers a state-government operating model for exploring broad knowledge-work use while documenting the adoption barriers that can keep a nominally available tool from becoming routine practice.","evidence":"Participants estimated saving 95 minutes per day and most described the experience as very positive. The report also found no single successful-user profile and documented inaccuracy, habit formation, lack of learning time, a steep learning curve, and privacy uncertainty as material barriers.","architectureImplications":"Provide an approved enterprise environment, but pair it with a role-based use-case library, safe input examples, output-review steps, and telemetry that can test self-reported savings against observable workflow measures.","governanceImplications":"Use embedded AI ambassadors, communities of practice, simple dos and don'ts, human ownership of work products, and repeated user research before scaling to higher-risk or team-based workflows.","securityPrivacyImplications":"Participants remained uncertain about how inputs were processed and stored despite enterprise terms and existing policy, showing that contractual protection must be translated into plain, scenario-specific guidance.","caveats":"The evaluation was a volunteer pilot, not a controlled study; the 95-minute estimate was self-reported, 136 of 175 participants provided direct feedback, and Carnegie Mellon supported the effort consultatively rather than acting as an independent evaluator."}},{"version":"enrichment/2026-09-05T02:33:27.019Z/pennsylvania-chatgpt-pilot","resource":{"id":"pennsylvania-chatgpt-pilot","title":"State workforce pilot reports large perceived savings but uneven readiness","organization":"Commonwealth of Pennsylvania","sector":"State government","geography":"Pennsylvania, United States","publishedAt":"March 2025","publicationDate":null,"eventDate":null,"sourceName":"Lessons from Pennsylvania's Generative AI Pilot with ChatGPT","sourceLabel":"Commonwealth pilot report","sourceUrl":"https://www.pa.gov/content/dam/copapwp-pagov/en/oa/documents/programs/information-technology/documents/openai-pilot-report-2025.pdf","evidenceClass":"government-evaluation","outcomeClass":"mixed","topics":["knowledge-work","data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"Pennsylvania equipped 175 employees across 14 agencies with ChatGPT Enterprise for a yearlong pilot using surveys, focus groups, office hours, and role-specific support.","sledRelevance":"The pilot offers a state-government operating model for exploring broad knowledge-work use while documenting the adoption barriers that can keep a nominally available tool from becoming routine practice.","evidence":"Participants estimated saving 95 minutes per day and most described the experience as very positive. The report also found no single successful-user profile and documented inaccuracy, habit formation, lack of learning time, a steep learning curve, and privacy uncertainty as material barriers.","architectureImplications":"Provide an approved enterprise environment, but pair it with a role-based use-case library, safe input examples, output-review steps, and telemetry that can test self-reported savings against observable workflow measures.","governanceImplications":"Use embedded AI ambassadors, communities of practice, simple dos and don'ts, human ownership of work products, and repeated user research before scaling to higher-risk or team-based workflows.","securityPrivacyImplications":"Participants remained uncertain about how inputs were processed and stored despite enterprise terms and existing policy, showing that contractual protection must be translated into plain, scenario-specific guidance.","caveats":"The evaluation was a volunteer pilot, not a controlled study; the 95-minute estimate was self-reported, 136 of 175 participants provided direct feedback, and Carnegie Mellon supported the effort consultatively rather than acting as an independent evaluator.","streamIds":["state-government"],"roles":{"sales":"Interpretation — Customer problem: agencies may provide an enterprise assistant while staff remain unsure how to use it safely or make it part of routine work. Stakeholders: innovation and program leaders, workforce development, IT, privacy, managers, and prospective ambassadors. Discovery: which tasks recur; what learning time is available; where do employees misunderstand input processing; and who reviews generated work? Value hypothesis: role-specific support may improve useful adoption and reveal measurable workflow gains. Potential engagement: a use-case and readiness workshop followed by supported pilot cohorts. Unsupported claims: participants' estimated 95 minutes saved per day is not a validated forecast, and the volunteer pilot does not establish causal productivity gains or independent evaluation by Carnegie Mellon.","engineering":"Interpretation — Fit: begin with approved knowledge-work tasks in an enterprise environment, retaining human review of outputs. Architecture and integration: pair access with a role-based use-case library, scenario-specific safe inputs, and workflow telemetry; do not assume deeper integrations are required. Prerequisites: confirmed input-processing and retention terms, accessible guidance, baseline tasks, and an owner for output quality. Constraints: inaccuracies, habit formation, and learning demands can limit value despite positive sentiment. Security: test the approved handling rules against realistic sensitive-input scenarios and explain where prompts and outputs are stored or processed. Proposed proof: compare observable cycle time and reviewed quality with participant estimates across different roles, tracking unsafe-input confusion and usability barriers separately.","delivery":"Interpretation — Work: recruit agency ambassadors, run office hours and communities of practice, provide practice time, and maintain a reviewed use-case library. Dependencies: manager support, privacy answers in plain language, and enough baseline evidence to assess actual workflow change. Ownership: program managers own work products; an adoption lead coordinates support; IT/privacy teams maintain environment and handling guidance. Skills and adoption: teach task selection, output checking, and safe input through role-specific examples instead of relying on a generic policy. Governance checkpoints: onboarding readiness, cohort feedback, and review before higher-risk or team-based use. Proposed acceptance: participants demonstrate approved handling, use targeted workflows routinely, and document observed quality/time changes. Risks include self-report inflation, inadequate learning time, and persistent privacy uncertainty despite enterprise terms."},"retrievedAt":null,"enrichedAt":"2026-09-05T02:33:27.019Z","enrichmentBasis":"archived evidence"}}]}