Pennsylvania's new chief data and AI officer is spending her first months on the job building the infrastructure needed to make government services easier to use - and she says the technology is only part of the equation.
Priya Narasimhan started June 1 as the state's inaugural chief data and AI officer, a role that combines data and artificial intelligence responsibilities under one leader for the first time. Her approach to the job comes down to a phrase she repeats to her teams: "fall in love with the problem."
"Maybe the right answer is not AI," Narasimhan said. "Maybe it is some other technology, or it could simply be a process change or, you know, just connecting some people and processes together."
Why the role was created
Pennsylvania previously had a chief data officer, but the new position formally merges data and AI oversight. The reasoning, Narasimhan said, is that AI and data have become too intertwined to manage separately. Responsible AI depends on well-governed, high-quality data underneath it, and combining the functions gives the state a better chance to treat them as one effort rather than building siloed strategies.
The role builds on Gov. Josh Shapiro's 2023 executive order establishing principles for AI adoption. Narasimhan's responsibilities include setting statewide direction, strengthening data governance and helping agencies adopt AI where it can improve service delivery.
Narasimhan came to the role from Mathematica, where she worked as a federal contractor. That experience gave her a firsthand look at how government programs, procurement and service delivery operate. She entered public-sector technology during the COVID-19 era, when problems with unemployment compensation systems made the stakes of government technology especially clear.
"That made it very real for me as a technologist," she said. "I saw government technology and the systems that are driving these services, and how much they matter in people's lives."
Where data and AI fit
Narasimhan sees her office as a connective layer across Pennsylvania's technology operations, one that has to align with the broader IT organization. Modern data architecture is central to that work, particularly where private-sector approaches could help agencies make better use of their information.
That doesn't mean building one giant database of everything Pennsylvania knows. Agencies can remain stewards of the data they know best while the enterprise provides common standards, infrastructure and secure connections that make sharing and using information easier when appropriate.
Her first-year priorities reflect that foundation-building: developing a statewide data and AI strategy, testing it through practical initiatives that can improve services and be reused elsewhere, and building the teams and operating model needed to sustain the work. For government professionals looking to build similar capabilities, resources like the AI Learning Path for Policy Makers offer structured guidance on governance and implementation.
Some of the AI work already being explored in Pennsylvania is decidedly practical. During the state's response to HR1, the One Big Beautiful Bill Act, residents submitted blurry or incorrect documents. AI could help identify those problems earlier, giving people a chance to correct them before delays occur.
Narasimhan also sees opportunity on the technology side, particularly around legacy systems. AI could help engineers make sense of complicated code before migration, potentially cutting down on time and reducing the technical risk of moving older systems. Agencies are also looking at document processing and knowledge search, including ways to help employees sort through large amounts of information and make government websites easier for residents to navigate.
When AI decisions carry more weight
Narasimhan is also thinking carefully about where AI belongs - and where it needs closer scrutiny. The stakes rise when technology moves into decisions involving programs and benefits, she said, which is why Pennsylvania is paying close attention to privacy, bias and transparency. That includes being able to explain how an AI system reaches a conclusion and keeping people involved in important decisions.
The same thinking extends to procurement. AI is increasingly embedded in everyday software, Narasimhan noted, which means the state may not always know it is buying an AI capability just by reading a product label. Her office is working with procurement, governance and cybersecurity teams on stronger expectations for vendors around data use, transparency and chain of custody. The state is also exploring a risk-based review process, where lower-risk uses move quickly while higher-risk applications get more oversight. AI for Government covers similar ground for public-sector teams working through adoption and governance questions.
Why this matters for government professionals
For Narasimhan, the real test of this work will happen far from strategy meetings and governance frameworks. It will happen when someone interacts with Pennsylvania government and finds that something simply works a little better.
"A year from now, the citizens should experience it in their service delivery," she said. "That is the real proof that we want to work towards."
Residents may not notice the technology itself. They may simply find a government website easier to navigate, a document process simpler, or an answer without digging through pages of information. For government professionals, the lesson is that AI success depends less on flashy applications and more on the unglamorous work underneath: data standards, governance, procurement practices and infrastructure. Those are the pieces that determine whether the technology actually improves services - or just adds another layer of complexity.
Your membership also unlocks:
