Pennsylvania Expands Generative AI for State Employees with Copilot and ChatGPT Enterprise

Pennsylvania is moving from pilot to purchase order: Governor Josh Shapiro told more than 900 technology, academic and business leaders at the AI Horizons Summit in Pittsburgh that the commonwealth will expand access to advanced generative AI tools for qualified state employees — adding Microsoft Copilot to an existing ChatGPT Enterprise rollout and wrapping the expansion with training, oversight and economic development commitments.

Background​

The Shapiro administration’s effort builds on a year-long, state-run pilot with OpenAI’s ChatGPT Enterprise that began in January 2024. That pilot — run by the Office of Administration in partnership with Carnegie Mellon University and OpenAI — involved roughly 175 employees across 14 agencies and produced headline metrics the administration now uses to justify wider deployment. Participants reported large perceived time savings and broadly positive experiences during the pilot.
Pennsylvania’s announcement at the AI Horizons Summit frames the work as part of a three‑pronged strategy: increase government productivity, protect citizen data and build the state’s AI economy through public‑private partnerships and workforce training. The administration also points to an independent assessment that ranks Pennsylvania among the top three U.S. states for AI readiness.

---

What the state announced at AI Horizons​

• Commonwealth employees who qualify will be given access to a vetted suite of generative AI tools: continuing ChatGPT Enterprise access and adding Microsoft Copilot Chat as part of the new, expanded offering. The administration called this “the most advanced suite of generative AI tools offered by any state,” a characterization rooted in its dual‑vendor approach.
• The rollout will be accompanied by official governance structures: the continuation of the Generative AI Governing Board (established by Executive Order 2023‑19), the creation of a Generative AI Labor and Management Collaboration Group to involve unions and workers, and mandatory training for employees who use the tools.
• The summit also included new industry and academic commitments: a five‑year, $10 million research partnership between BNY (Bank of New York) and Carnegie Mellon University to found the BNY AI Lab focused on governance and accountability, and a Google‑run AI Accelerator aimed at bringing free training and tools to Pennsylvania small businesses.

These moves represent a shift from experimental pilots toward a managed, enterprise‑grade deployment model with explicit workforce engagement and external research funding. The administration positions the work both as operational modernization and as economic development.

---

The pilots and the numbers: what happened, and how to read the claims​

The “95 minutes a day” headline​

The most attention‑grabbing figure from Pennsylvania’s pilot is the reported average time savings: employees who used ChatGPT in the pilot said they saved 95 minutes per day on tasks such as drafting emails, summarizing long documents, researching policy, and basic coding assistance. That figure comes from the state’s exit surveys, interviews and structured feedback collected during the pilot and has been repeated by the governor, Carnegie Mellon University and state press materials.
Independent public‑sector reporting also examined the pilot methodology and results: coverage in industry outlets notes the pilot’s mixed methods (surveys, telemetry, interviews) and highlights that while reported time savings were large, outputs frequently required human verification and editing — a common reality for today’s generative models. In short, the 95‑minute figure is a self‑reported, pilot‑derived metric that indicates substantial perceived productivity gains, but it does not substitute for an independent, audit‑level productivity study across all state roles.

How the administration’s economic claims compare​

At the summit, the administration reiterated that its economic strategy — which includes AI and energy investments — has helped the state attract major private commitments. The administration’s materials around the event cite more than $25 billion in private‑sector commitments and roughly 11,000–12,400 new jobs since the governor took office, though the exact totals differ slightly across press releases and departmental pages. Some official pages list the total as about $25.2 billion and nearly 11,000 jobs, while other briefings and summit materials cite $25.6 billion and 12,400 jobs. Those differences appear to reflect fast‑moving announcements and aggregated reporting windows; readers and procurement officials should treat headline totals as rolling figures that are best verified against the Department of Community & Economic Development’s project database for any specific project or job claim.

“Most advanced suite of generative AI tools offered by any state” — a state claim, not an independent rating​

The administration’s characterization of Pennsylvania’s toolset as the “most advanced” among U.S. states is an aspirational, comparative claim based on the state running both ChatGPT Enterprise and Microsoft Copilot in government contexts. That phrasing is best understood as a promotional positioning: there is no single, objective checklist published by an independent authority that ranks every state by the exact combination of vendor products, tenancy models, governance guardrails and training programs. The claim therefore stands as a defensible marketing statement by the commonwealth, but one that should be treated as the state’s self‑assessment rather than an independently audited fact.

---

Technical and procurement details: what IT leaders need to know​

Which products, and what they mean in practice​

• ChatGPT Enterprise: a commercial OpenAI offering that provides tighter administration controls, stronger data protections, and enterprise management features compared with public consumer accounts. Pennsylvania’s pilot used ChatGPT Enterprise to limit training reuse of state data and to apply administrative controls over access and usage.
• Microsoft Copilot Chat / Microsoft 365 Copilot: an assistant integrated across Office apps that can summarize email threads, draft documents, generate slides and automate repetitive tasks inside Word, Outlook, PowerPoint, Excel and Teams. Deploying Copilot in government environments typically involves Microsoft’s secure tenancy options (including Azure Government or GCC equivalents) and enterprise governance controls like Purview classification, Data Loss Prevention policies and audit logging. The addition of Copilot gives state staff a productivity assistant that is deeply integrated into the Microsoft 365 workflow many agencies already use.

Both product classes emphasize “human‑in‑the‑loop” usage: models assist with drafting and analysis, but final decisions and official documents remain the responsibility of trained employees and reviewers. That approach is consistent with the executive order’s principles for accuracy, transparency, privacy and human oversight.

Security and compliance checklist (practical)​

1. Classify data and apply sensitivity labels before you permit AI access.
2. Route high‑sensitivity and controlled unclassified information (CUI) only through cleared tenancy (e.g., Azure Government or equivalent).
3. Enable robust audit logs, retention policies and eDiscovery to support transparency and FOIA responses.
4. Deploy least‑privilege access and phishing‑resistant MFA for accounts that can prompt Copilot or ChatGPT.
5. Require prompt provenance logs and mandate human verification steps for legal, benefits, licensing or safety‑critical outputs.

These measures are consistent with best practices seen in other public‑sector pilots and federal guidance: tool integration must be combined with data classification, DLP enforcement and well‑documented review processes to limit risk.

---

Governance, labor and oversight: the administration’s approach​

The Shapiro administration is explicit about worker involvement. The new Generative AI Labor and Management Collaboration Group is designed to give unions and employees a formal voice in how AI is introduced across roles, an arrangement intended to reduce resistance and to design augmentative workflows rather than wholesale replacements. That aligns with the governor’s public line that AI is a “job enhancer, not a job replacer.”
Governance also rests on the existing Executive Order (2023‑19) that established a Generative AI Governing Board and codified principles of accuracy, privacy, equity and transparency for state AI use. The governing board is responsible for policy, vetting vendor contracts, and approving expansion plans — a model that puts central control in the Office of Administration while allowing agencies to pilot specialized use cases under established guardrails.
This hybrid approach — strong central policy plus worker collaboration and agency‑level pilots — is increasingly common among states that have moved beyond exploratory skunkworks into enterprise deployments. Independent assessments have rewarded states that pair policy with capacity building, and Code for America’s Government AI Landscape Assessment highlighted Pennsylvania as an “advanced” state in leadership and capacity building.

---

Partnerships and ecosystem building​

BNY‑CMU: $10 million for an AI lab focused on governance​

BNY and Carnegie Mellon announced a five‑year, $10 million collaboration to establish the BNY AI Lab at CMU’s School of Computer Science. The lab will concentrate on governance, trust and accountability for mission‑critical AI — an investment that both advances academic research and creates a local pipeline of applied expertise for financial and government systems. The announcement was carried by both the institution and national press outlets.

Google’s AI work and the small‑business accelerator​

Google announced a statewide accelerator and training effort for Pennsylvania small businesses tied to the summit, part of a broader commitment to workforce and infrastructure investment in the region. Google’s outreach materials and state press briefings describe free training and toolkits aimed at helping entrepreneurs reduce costs and scale operations with AI: a classic public‑private skills initiative that couples vendor expertise with SME support.
Taken together, these partnerships — academic, corporate and governmental — form the scaffolding for a regional AI cluster: research funding, skilling programs and vendor partnerships that both accelerate public deployments and build private‑sector opportunity around the state.

---

Risks, limitations and the accountability imperative​

No public‑sector AI rollout is risk‑free. The Pennsylvania pilot and subsequent public commentary surfaced several recurring concerns that any state must manage proactively:

• Accuracy and hallucinations: generative models can make confident but incorrect assertions. Human verification is non‑negotiable for legal, medical, or benefits decisions. The pilot explicitly emphasized the need for human oversight and additional verification steps.
• Privacy, FOIA and data residency: inputs and outputs may be subject to public records laws. Contracts must explicitly define retention, exportability, training reuse restrictions and vendor obligations for FOIA responses. Agencies must classify data before permitting any AI interaction.
• Vendor lock‑in and portability: heavy dependence on a single cloud or assistant risks long‑term lock‑in. Procurement should require data egress clauses, measurable SLAs and clear audit rights. Federal and municipal pilots have repeatedly underscored this as a governance priority.
• Equity and bias: models trained on imbalanced data can produce biased outputs. Regular fairness testing, diverse red‑team reviews and publicly reported audit results are needed to maintain trust.
• Workforce impacts and reskilling: while the administration frames AI as augmentative, role redesign will be required. The Labor and Management Collaboration Group is a step toward equitable transition, but robust retraining, redeployment plans and measurable outcomes will be necessary to keep the promise.

These hazards are not hypothetical; they are the operational realities seen in federal and international pilots. Good governance is therefore not optional — it’s the central variable between productive modernization and a public relations misstep.

---

Practical steps for state and local IT teams​

For CIOs, procurement leads and digital services teams planning similar rollouts, the Pennsylvania experience offers a practical checklist:

1. Start with a short, instrumented PoV (proof of value) focused on a few high‑impact, low‑risk workflows.
2. Document baseline metrics (AHT, throughput, error rates) so claimed savings can be quantified rather than self‑reported alone.
3. Mandate human‑review thresholds and require versioned prompt logs for auditability.
4. Build training programs with clear competency goals (prompt engineering, verification, privacy hygiene).
5. Negotiate contracts with portability, audit rights, and explicit non‑training clauses if you cannot allow vendor retraining on sensitive data.
6. Coordinate with unions and human resources to design role‑redesign pathways and reallocation of saved capacity to higher‑value public services.

These steps mirror recommendations in federal and state evaluations and respond directly to common pitfalls seen in other Copilot and generative AI pilots.

---

What to watch next​

• Execution: moving from pilot to broad deployment is an operational challenge. The state must deliver on training, DLP enforcement, tenancy configuration, and centralized auditing to make the admin’s claims credible at scale.
• Measured outcomes: independent third‑party audits and open, repeatable metrics will be crucial. The 95‑minute figure is compelling, but wider adoption warrants longitudinal measurement across functions, not only exit surveys.
• Procurement posture: watch contract language for portability and data‑use limitations. States that swallow commercial convenience risk longer‑term constraints on policy, cost and sovereignty.
• Public transparency: to sustain public trust, the state should publish red‑team results, governance minutes and annual transparency reports detailing deployments, incidents, and outcomes.

---

Conclusion​

Pennsylvania’s announcement at the AI Horizons Summit is a significant, carefully staged example of how a U.S. state can move from experimentation to enterprise adoption with generative AI. The administration paired tools (ChatGPT Enterprise and Microsoft Copilot), governance (executive order, governing board, labor collaboration), and ecosystem investments (BNY‑CMU lab, Google training programs) to create a coherent narrative about productivity and economic growth.
The pilot’s headline metric — an average reported time savings of 95 minutes per day — is impressive and supported by state and university reporting, but it should be interpreted as pilot‑level, self‑reported evidence rather than a definitive audit of system‑wide productivity. The administration’s broader economic and readiness claims are backed by third‑party assessments (Code for America) and multiple corporate commitments, but some headline economic totals vary across official pages and should be verified at the project level for procurement and budgeting decisions.
For other public organizations watching closely: Pennsylvania’s playbook offers a balanced path — combine focused pilots, workforce engagement, robust procurement safeguards and independent research partnerships — but success will be measured in disciplined execution, transparent metrics and relentless attention to privacy, equity and oversight.

---

Source: fox43.com https://www.fox43.com/article/news/local/shapiro-pennsylvania-expands-generative-ai-tools-state-workers-ai-horizons-pittsburgh/521-49beaf97-1f77-41f2-bdbf-b3c0d3b33519/

 

[ChatGPT]

Pennsylvania’s government is moving from pilot projects to enterprise adoption of generative AI, announcing a statewide expansion that will give qualified state employees access to ChatGPT Enterprise and Microsoft Copilot alongside a governance and training regimen designed to manage risk and capture productivity gains.

Background​

Pennsylvania’s announcement at the AI Horizons Summit in Pittsburgh follows a year‑long pilot with OpenAI’s ChatGPT Enterprise and a public rollout strategy that pairs vendor products with oversight mechanisms. The pilot—run by the Office of Administration in partnership with Carnegie Mellon University and OpenAI—involved roughly 175 employees across 14 agencies and produced headline metrics the administration now uses to justify wider deployment.
The administration frames the initiative as a three‑pronged strategy: increase government productivity, protect citizen data, and build the state’s AI economy through public‑private partnerships and workforce training. To support that narrative, Pennsylvania announced the addition of Microsoft Copilot to its existing ChatGPT Enterprise rollout and committed to governance structures intended to provide accountability and transparency.

What was announced at the AI Horizons Summit​

• Continued access for qualifying commonwealth employees to ChatGPT Enterprise and new access to Microsoft Copilot Chat, described by the administration as a dual‑vendor approach that creates “the most advanced suite of generative AI tools offered by any state.” That phrasing is an administration claim and should be read as promotional rather than an independently validated ranking.
• Extension of governance structures: continuation of the Generative AI Governing Board (established under Executive Order 2023‑19), creation of a Generative AI Labor and Management Collaboration Group, and mandatory training for employees authorized to use the tools.
• New ecosystem investments announced at the summit, including a five‑year, $10 million BNY–Carnegie Mellon collaboration to create the BNY AI Lab focused on governance and accountability, and a Google AI Accelerator offered to Pennsylvania small businesses providing training and tools.

These moves mark a deliberate transition from isolated pilots toward an enterprise‑grade deployment model that blends central policy with agency‑level use cases and worker engagement.

---

Why the governor argues this matters​

Governor Josh Shapiro presented the expansion as a way to speed up government services and fuel the state’s innovation economy. The administration pointed to the pilot’s reported productivity metric—an average time savings of 95 minutes per day per participant—as evidence of measurable gains, and to broader economic claims that the administration has attracted more than $25 billion in private‑sector commitments since taking office. Both figures were used to justify the scaling decision.
It is important to treat these numbers with clear analytical caution. The 95‑minute figure is derived from the state’s exit surveys, interviews and internal pilot feedback—not from an independent, audit‑level productivity study. That means it signals strong perceived benefits among pilot participants but does not, on its own, quantify net productivity outcomes across diverse job classes or measure downstream effects such as error rates, rework, or changes to decision quality.

---

Overview of the technical and procurement posture​

What the two products represent in practice​

• ChatGPT Enterprise: an OpenAI commercial product configured for enterprises with administrative controls and contractual commitments around data use. In government pilots, commercial enterprise licensing typically includes features to limit vendor model‑training on customer data and offers administrative management and logging capabilities.
• Microsoft Copilot / Microsoft 365 Copilot: an assistant embedded across Office apps (Word, Outlook, PowerPoint, Excel, Teams) that can summarize threads, draft documents, and automate repetitive tasks. Deployments in public sectors typically use Microsoft’s secure tenancy options (e.g., Azure Government or GCC variants), Purview classification and data loss prevention (DLP) policies to control information flow. Copilot’s tight integration into existing Microsoft 365 workflows is an operational reason many agencies choose it.

Combining both products gives staff multiple toolsets depending on workflow: a conversational RAG (retrieval‑augmented generation) assistant and an app‑embedded productivity assistant. That dual approach is the administration’s stated basis for calling it the most advanced state offering, though it is best treated as a strategic choice rather than an objective external ranking.

Procurement and tenancy considerations​

Deploying Copilot or ChatGPT Enterprise in a state context is not a simple license purchase. Practical technical requirements include:

• Configuring secure tenancies (Azure Government, GCC, or equivalent) to ensure appropriate data residency and compliance.
• Implementing data classification so that Personally Identifiable Information (PII), Controlled Unclassified Information (CUI), and other sensitive records are routed only through cleared environments.
• Enabling auditing, retention and eDiscovery to support transparency and public records (FOIA) obligations.
• Applying least‑privilege access and phishing‑resistant multi‑factor authentication (MFA) for accounts that can prompt AI assistants.

These technical controls are recurring recommendations from federal and state pilots, and Pennsylvania’s plan explicitly acknowledges the need for them.

---

Governance, labor and oversight: the state’s approach​

Pennsylvania is attempting a governance model that mixes central policy with worker participation:

• The Generative AI Governing Board oversees policy, vendor vetting, and expansion approvals under the executive order issued in 2023. That central body is meant to ensure baseline standards across agencies.
• The newly formed Generative AI Labor and Management Collaboration Group aims to bring unions and employees into the implementation design to shape role changes and training, reducing the risk of one‑sided automation.
• Mandatory training and competency requirements for employees who will use these tools are a stated condition of access, paired with human‑in‑the‑loop mandates for high‑risk outputs.

This hybrid governance posture — central guardrails plus agency innovation and worker input — mirrors best practices recommended in independent assessments of government AI pilots. But governance on paper must be matched by rigorous operational execution: training completion rates, audit trails, red‑team results, and published accountability metrics will determine long‑term credibility.

---

The economic and research angle: building an AI cluster​

At the summit Pennsylvania highlighted ecosystem building as part of the strategy:

• BNY–Carnegie Mellon collaboration: a five‑year, $10 million commitment to establish the BNY AI Lab at CMU’s School of Computer Science, focused on governance, trust and accountability in mission‑critical systems. The lab aims to create applied research that supports both the finance and public sectors.
• Google AI Accelerator for small businesses: announced as free training and tool access to help Pennsylvania entrepreneurs streamline operations and reduce costs through AI. Such programs are classic public‑private workforce development plays that can help diffuse capability beyond government.

These investments aim to anchor talent and technical expertise in the region, boosting the state’s claim of being among the top U.S. states for AI readiness. Still, tracking real outcomes—number of researchers trained, patents filed, small business cost savings, job quality changes—will show whether the investments create long‑term cluster effects or simply short‑term visibility.

---

Strengths of Pennsylvania’s plan​

• Coherent layering of tools and governance: pairing ChatGPT Enterprise with Microsoft Copilot while retaining a governing board and labor collaboration structure reflects a balanced move from experimentation to managed scale.
• Practical focus on workforce readiness: mandatory training and worker engagement acknowledge that technological adoption must be accompanied by reskilling and role redesign.
• Ecosystem investments: the BNY AI Lab and Google accelerator link public deployments to academic research and small business skilling, a useful tactic to grow local capacity.
• Technical defensibility: emphasis on secure tenancy options, DLP, Purview classification and audit logging shows awareness of compliance realities when AI touches public data.

These strengths make the policy posture defensible for a government looking to modernize back‑office workflows while retaining oversight.

---

Risks, limitations and what to watch closely​

1. The 95‑minute claim is headline‑worthy but built on internal metrics​

The widely cited figure that pilot participants saved 95 minutes per day is compelling but is a self‑reported metric derived from exit surveys and structured feedback. Self‑reported time savings often overstate actual net gains because they do not always capture verification time, error correction, or the cognitive overhead of supervising AI outputs. Independent, longitudinal audits and baseline measurements (AHT, error rates, throughput) are needed before scaling claims to the whole workforce.

2. Accuracy, hallucinations and verification requirements​

Generative models can produce confident but incorrect outputs. For legal, benefits, licensing, or health‑related tasks, human verification is non‑negotiable. The administration’s materials emphasize human‑in‑the‑loop usage; success depends on enforcing that practice operationally, not just via policy statements.

3. Public records, FOIA and data residency issues​

Inputs and outputs may be subject to public records laws. Contracts must clearly define retention and exportability, and procurement should include non‑training clauses or data‑use restrictions if the state cannot permit vendor model retraining on sensitive data. Without these protections, the state risks complicated FOIA responses and potential loss of control over sensitive information.

4. Vendor lock‑in and procurement posture​

A dual‑vendor approach mitigates single‑vendor dependency but does not eliminate lock‑in risks. Procurement language should include explicit egress clauses, audit rights, and SLAs for portability to avoid long‑term technical and fiscal entanglements.

5. Workforce impacts and job redesign gaps​

Even with labor collaboration groups, role redesign and reskilling are hard to operationalize at scale. The administration must define measurable outcomes for retraining, redeployment, and job quality to ensure saved hours translate into higher‑value public service rather than hidden layoffs or flattened career ladders.

6. Transparency and independent evaluation​

To maintain public trust, Pennsylvania should publish red‑team results, independent audits, and annual transparency reports detailing deployments, incidents, and measurable outcomes. Without independent verification, productivity and economic claims risk appearing promotional rather than evidentiary.

---

Practical checklist for state and local IT leaders​

For IT leaders planning similar deployments, Pennsylvania’s approach suggests a checklist that others can adapt.

1. Establish a governing body that vets vendor contracts and approves expansions.
2. Start with instrumented proofs‑of‑value (PoVs) that document baseline metrics (AHT, throughput, error rates).
3. Classify data and apply sensitivity labels before enabling AI access.
4. Route high‑sensitivity and CUI only through cleared tenancy (Azure Government or equivalent).
5. Require least‑privilege access, phishing‑resistant MFA, and prompt provenance logs.
6. Mandate human verification thresholds for any legal, health, benefits or safety‑critical outputs.
7. Negotiate procurement clauses for portability, non‑training of vendor models on state data, and audit rights.
8. Build role‑based training programs with clear competency markers and workforce transition plans.

These steps mirror both the Pennsylvania playbook and recommended federal best practices.

---

Recommendations for making the rollout credible and durable​

• Publish an independent third‑party audit plan to validate pilot claims and track longitudinal outcomes across agencies.
• Mandate public transparency: publish the governing board’s meeting minutes, red‑team results, and annual deployment and incident reports.
• Tie funding for the next expansion phase to measurable milestones: training completion rates, audit logs, FOIA responsiveness, error reduction and demonstrable net time savings after verification time is included.
• Use the BNY–CMU lab to operationalize governance research into applied audits and tooling for bias detection, fairness testing and provenance logging.
• Implement pilot‑stage KPIs that include quality and not just speed: measure rework, error remediation time and downstream citizen outcomes.

These steps reduce the chance that a promising early narrative becomes a cautionary tale.

---

What the announcement means for Windows‑centric IT professions​

For system administrators, desktop engineers and IT managers who operate Windows environments across public agencies, several tangible implications arise:

• Expect increased demand for identity and access governance skills (MFA, conditional access, least‑privilege models) as administrators control who can call AI services and what data they can access.
• Expect more integration work around Purview classification, DLP rules and audit logging to ensure that Copilot and ChatGPT operate only on appropriately labeled content.
• Tooling and endpoint management will need to standardize prompt provenance logging and secure tenancy configurations to support FOIA and eDiscovery requests.
• Training and change management will be essential—technical rollout without user competency programs will amplify operational risk.

In short, IT teams should prepare for a heavier emphasis on governance, telemetry and documentation than they face with typical desktop deployments.

---

Conclusion​

Pennsylvania’s move to expand access to ChatGPT Enterprise and Microsoft Copilot represents a deliberate, well‑scaffolded shift from pilot to enterprise adoption. The administration married technical controls, worker engagement and ecosystem investments in a way that makes sense for a state seeking both operational modernization and economic growth.That said, the most important work starts now: converting self‑reported pilot gains into independently verifiable outcomes, operationalizing human‑in‑the‑loop controls across hundreds or thousands of workers, and ensuring procurement language protects data sovereignty and FOIA responsiveness. The 95‑minute figure and “most advanced suite” claim are useful for narrative, but they are not substitutes for transparent metrics and independent evaluation. If Pennsylvania follows through with audits, published governance outputs, rigorous procurement safeguards, and measurable workforce transition plans, the state will have a strong case study for responsible, productive public‑sector AI adoption.

---

Source: WNEP https://www.wnep.com/article/news/state/shapiro-pennsylvania-expands-generative-ai-tools-state-workers-ai-horizons-pittsburgh/521-49beaf97-1f77-41f2-bdbf-b3c0d3b33519/

 

[ChatGPT]

Pennsylvania’s decision to expand access to enterprise-grade generative AI for qualified state employees — pairing ChatGPT Enterprise with Microsoft Copilot Chat and a governance framework — marks one of the most ambitious state-level AI rollouts in the United States and moves the commonwealth from pilot experimentation toward enterprise operationalization.

Background​

Pennsylvania’s announcement at the AI Horizons Summit in Pittsburgh builds on a year-long pilot of ChatGPT Enterprise run by the Office of Administration in partnership with Carnegie Mellon University and OpenAI. The pilot involved roughly 175 employees across 14 agencies and produced headline metrics the administration now cites to justify wider deployment.
Governor Josh Shapiro framed the expansion as a three‑pronged strategy: increase government productivity, protect citizen data, and seed an AI-powered economic ecosystem anchored by local institutions and private partners. The administration characterized the combined offering as “the most advanced suite of generative AI tools offered by any state,” a promotional claim that the administration itself and independent observers caution should be treated as a strategic positioning rather than an independently audited ranking.

---

What the announcement actually changes​

Tools and coverage​

• Qualified commonwealth employees will retain access to ChatGPT Enterprise and will also be offered Microsoft Copilot Chat as part of an expanded, dual‑vendor approach. The administration’s stated rationale is that the two tools serve complementary workflows: ChatGPT Enterprise as a flexible conversational assistant and Copilot Chat as a deeply integrated assistant inside Microsoft 365 apps.
• The move is not merely about provisioning seats; it includes mandatory training, governance board continuations, worker engagement via a new Generative AI Labor and Management Collaboration Group, and investments designed to anchor research and small-business support in Pennsylvania.

Pilot outcomes that influenced the decision​

The most-cited metric from the pilot is a reported average time savings of 95 minutes per day by participants for tasks like drafting emails, summarizing documents, researching policy, and simple coding assistance. That figure comes from exit surveys, interviews, and structured feedback collected during the pilot and has been repeated in state briefings. Observers and coverage of the pilot note this is a self‑reported metric captured in a limited cohort; it signals perceived productivity gains but is not an independent, audit‑level measure of net productivity across the full workforce.

---

Why Pennsylvania’s approach matters​

A dual‑vendor playbook​

By offering both ChatGPT Enterprise and Microsoft Copilot Chat, Pennsylvania is betting on a multi-tool strategy that recognizes different integration points and governance postures:

• ChatGPT Enterprise provides conversational RAG-style workflows and enterprise administrative controls that allow centralized logging and contractual restrictions on vendor model training on state data.
• Microsoft Copilot Chat integrates directly into Word, Outlook, PowerPoint, Excel, and Teams, making it operationally attractive to staff already embedded in Microsoft 365 workflows. Copilot deployments typically rely on Microsoft enterprise tenancy protections (e.g., Azure Government / GCC) and Purview/DLP controls.

This dual approach gives staff options: a broader conversational assistant and a tightly integrated productivity assistant. That choice is persuasive for operational leaders who must balance utility, vendor lock‑in risk, and the technical burden of integrating new services into existing identity, data classification, and DLP frameworks.

Governance and worker involvement​

Pennsylvania explicitly layered governance and labor collaboration into the expansion:

• The Generative AI Governing Board (established by Executive Order 2023‑19) remains the central policy body approving expansions, vendor vetting, and high-level guardrails.
• The newly created Generative AI Labor and Management Collaboration Group is designed to involve unions and front-line workers in shaping deployments and role-based guardrails — a practical step to reduce resistance and design augmentative workflows rather than one-size-fits-all automation.

This combination addresses two common public-sector failure modes: technology-first procurement without worker buy-in, and policy frameworks that are too abstract to govern day-to-day use. Having labor at the table and a governing board with documented principles (accuracy, privacy, equity, transparency) is foundational to durable deployments.

Training and workforce readiness​

The administration reports robust training uptake: more than 1,300 employees have completed InnovateUS training on safe and ethical AI use, with another 3,200 enrolled. These numbers indicate an intent to scale competency development alongside the tools themselves, not after problems surface. Training is a key determinant of whether AI becomes an augmentation layer—or a compliance and error generator.

---

Economic and ecosystem commitments​

Pennsylvania paired the operational announcement with ecosystem investments meant to turn a pilot into a regional cluster:

• A five‑year, $10 million research partnership between BNY and Carnegie Mellon University to create the BNY AI Lab, focused on governance and accountability. This aims to convert research into operational tools for auditability and testing of mission‑critical systems.
• A Google-run AI Accelerator aimed at small businesses to provide free training and tooling — a classic public‑private skills initiative to broaden benefits beyond government.

These investments signal a broader economic strategy: attract vendor commitments, create applied research capacity locally, and build a training pipeline that supports both public operations and private-sector growth. The administration has paired this with claims of attracting hundreds of millions (and in aggregate tens of billions) of private-sector commitments since the governor took office — rolling figures that should be verified project-by-project for precise accounting.

---

Reading the numbers critically: strengths and caveats​

Notable strengths​

• Ambitious governance design: Combining a governing board, formal labor engagement, mandatory training, and academic partnerships creates a layered defense against typical rollout failures.
• Real-world pilot data: The pilot produced measurable user feedback and usage telemetry used to refine the approach — a stronger posture than many jurisdictions that rely solely on vendor promises.
• Local ecosystem alignment: Tying the rollout to CMU, BNY, and Google gives the state a local talent and research pipeline for governance, auditing, and applied AI work.

Major caveats and open questions​

• The “95 minutes/day” metric is self‑reported and cohort-limited. It signals perceived value but is not an independent, longitudinal measurement of net productivity gains across different job classes, nor does it account for verification time, rework, or error mitigation. Treat it as an encouraging pilot result, not an audited productivity delta.
• “Most advanced suite” is a promotional claim. There is no single authoritative ranking of states by the exact combination of vendor tools, tenancy types, governance scaffolds, and training programs. The administration’s phrasing is defensible as a self-assessment but should be distinguished from an independent benchmark.
• Data governance and FOIA remain complex. Even with enterprise contracts, public records laws, data residency and retention rules, and eDiscovery obligations create ongoing legal and operational obligations. The state must publish red-team results, independent audits, and regular transparency reports to maintain trust — steps that were recommended by observers and remain work-in-progress.
• Operational scale-up risk. Scaling from ~175 pilot participants to thousands of employees multiplies governance friction points: role-based access, DLP enforcement, incident response, and the human-in-the-loop controls that prevent automated errors from becoming systemic. The administration’s checklist for secure rollout (data classification, least-privilege, audit logging, MFA) will be essential to operational success.

---

Technical posture and operational checklist for IT leaders​

For IT teams and Windows-centric professionals implementing or auditing similar deployments, the Pennsylvania playbook highlights concrete technical responsibilities:

• Classify data and apply sensitivity labels before enabling AI access. Route Controlled Unclassified Information (CUI) and high-sensitivity content only through cleared tenancies (e.g., Azure Government).
• Extend Data Loss Prevention (DLP) policies and Purview classification to any flows that touch Copilot or ChatGPT. Ensure prompt provenance is logged.
• Enforce least-privilege and phishing‑resistant MFA for accounts authorized to use generative AI. Limit connectors for agents and require approvals for write operations.
• Standardize prompt-provenance logging, retention policies, and eDiscovery integration to support FOIA and audit requests.
• Negotiate procurement clauses that restrict vendor model training on state data, allow audit rights, and ensure portability where possible.
• Start with instrumented proofs-of-value (PoVs) that capture baseline metrics (average handling time, throughput, error rates).
• Require mandated human verification thresholds for any legal, medical, benefits, licensing, or safety‑critical outputs.
• Publish red-team results and independent audits to convert pilot claims into verifiable public metrics.

---

Labor, policy and public-trust implications​

Pennsylvania’s inclusion of a labor-management collaboration group is a pragmatic recognition that durable AI adoption depends on worker voice. In practical terms, this means:

• Negotiated role boundaries for where AI is permitted to draft, summarize, or recommend versus where final decision authority must remain with humans.
• Training and competency frameworks tied to role-based access, so only staff who demonstrate proficiency are allowed to use AI in production workflows.
• Transparency commitments — publishing meeting minutes, incident reports, and audit summaries — to maintain citizen trust and enable oversight. Observers recommended these items as critical for credibility.

Absent rigorous transparency and independent validation, the public narrative can swing from “AI saved time” to “AI introduced risk” very quickly; labor participation and open evidence are primary mitigants.

---

Recommendations and red‑flag checks for other states and agencies​

• Prioritize pilot integrity: instrument pilots so they capture not only speed gains but quality metrics (rework, corrections, error remediation time).
• Tie expansion funding to auditable milestones: training completion, SOC‑style logs, independent audits, FOIA responsiveness, and demonstrable net time saved after verification time is included.
• Use local academic partnerships strategically: fund lab work that converts red-team and bias-detection research into operational tooling that agencies can use. Pennsylvania’s BNY–CMU partnership provides a model for this.
• Avoid over-reliance on single vendor narratives; favor dual-path approaches that let agencies choose the best tool for the use case while central governance ensures compliance.

---

What remains to be seen​

• Will Pennsylvania publish independent, longitudinal audits that convert pilot feedback into validated productivity and quality metrics? Observers recommended this as a necessary step to maintain public trust.
• How will FOIA and records-retention obligations be operationalized at scale when prompts, responses, and agent actions are part of official decision trails? The administration has policy frameworks in place but translating those into automation-ready retention and export formats remains a heavy engineering task.
• Can the Generative AI Labor and Management Collaboration Group evolve beyond consultation into enforceable role definitions? Worker involvement is promising, but outcomes depend on contract language and implementation fidelity.

---

Conclusion​

Pennsylvania’s expanded rollout — combining ChatGPT Enterprise and Microsoft Copilot Chat with governance, training, and local research investments — is a bold, well-scaffolded step that other states will watch closely. The administration’s approach scores highly on governance design, worker engagement, and ecosystem-building, yet important questions remain about how pilot metrics translate into verifiable, enterprise-grade outcomes at scale.
The state’s headline figures — the reported 95 minutes of perceived daily time saved and the claim of offering the “most advanced suite” of tools — are powerful narratives that helped move policy from pilot to expansion. They should be read as pilot-derived and promotional respectively, not as final audit‑level proof. Converting promising pilot feedback into sustainable public value will require transparent audits, operationalized human‑in‑the‑loop controls, robust DLP and eDiscovery engineering, and continuous worker engagement.
For IT managers and Windows-focused practitioners, the takeaway is practical: plan for governance, telemetry, and identity-first controls before broad provisioning; require provenance and retention from day one; and ensure training and verification processes are baked into rollout milestones. Pennsylvania’s experiment is instructive — not as a finished blueprint — but as a high‑stakes case study in how states can move from AI hype to governed, labor‑informed operational practice.

---

Source: Pennsylvania Business Report - Pennsylvania state employees have most advanced AI tools in nation - Pennsylvania Business Report