How Pitt Built a Campus-Wide Strategy for Generative AI Tools

The Problem Every University Faces

When generative AI landed on campuses in early 2023, it didn’t wait for institutional approval. Students were already using it. Faculty were already worried about it. And administrators were already behind.

The typical response? Fragmented. A committee here, a policy draft there, a few brave professors redesigning syllabi alone at midnight.

Pitt’s challenge was the same as everyone else’s — but their approach to solving it was different enough to be worth studying.

The Tool Infrastructure: Equity First

Pitt’s institutional AI toolkit — Generative AI@Pitt — was built around two non-negotiable concerns: equity and privacy.

Mark Henderson, Pitt’s CIO, put it plainly: most students were already using free AI tools with zero privacy protections. By providing institution-approved platforms, Pitt ensures every student — regardless of financial situation — gets access to professional-grade tools with actual safeguards.

That’s a meaningful distinction. Free-tier AI tools often train on user inputs. Institutional agreements typically don’t. For a university handling sensitive research and student data, that gap matters enormously.

But Henderson was also careful to frame the tools correctly.

“The tools are just one piece. We’re not here to push a direction. We’re here to make sure everyone at Pitt has what they need to find their own.”

What Students Are Actually Doing With AI

Here’s where it gets interesting — and a little uncomfortable.

Last spring, faculty from four Pitt campuses ran GenAI Conversations, a study involving 95 students across focus groups. The headline finding: roughly 90% of Pitt students are already using generative AI in some form.

More revealing than the number was the texture of what students described. They talked about AI creating “fractures or tensions” in their relationships with faculty. They wondered how their own professors were using the same tools. They weren’t asking for permission — they were asking for honesty.

Annette Vee, who co-led the research and serves as faculty liaison for AI enablement at Pitt Digital, drew a clear conclusion: blanket prohibitions don’t work. The more productive move is redesigning courses with the reality of AI use already baked in.

That’s a harder ask than writing a policy. It requires faculty to actually engage — which is exactly what PASTA exists to support.

The Enablement Gap (And How Pitt Is Closing It)

Access to tools without the knowledge to use them well is just expensive shelfware.

Vee has spent two years running AI workshops for faculty in the Dietrich School of Arts and Sciences. Her observation is sharp: the coordination work happening across Pitt is broader and deeper than what any public-facing website yet reflects.

“Pitt Digital is really good at providing access to tools, but we also need to help faculty use them effectively. I think it’s crucial for faculty to know that they’re not on their own.”

The Center for Teaching and Learning and the Writing Institute both have resources on responsible AI integration. HAIL runs an ongoing speaker series and a biweekly newsletter. The DataSci+AI Forum drew nearly 400 attendees across two days in March and is now planned as an annual event.

The infrastructure for enablement is being built — it’s just not finished yet.

The Governance Playbook, Distilled

If you’re trying to build something similar at your institution, organization, or company, here’s what Pitt’s model actually demonstrates:

• Start with community, not compliance. PASTA existed before any formal committee. Trust and participation came first.
• Separate access from enablement. Deploying tools is step one. Teaching people to use them well is the actual work.
• Build formal and informal channels that talk to each other. The committee needs the lunch group. The lunch group needs the committee.
• Anchor everything in equity. Who gets access? Who gets left out? Those questions should drive infrastructure decisions, not follow them.
• Measure culture, not just adoption. Colaresi’s benchmark isn’t tools deployed or committees convened — it’s whether Pitt can model a fundamentally different relationship with AI.

The Bigger Takeaway

Pitt’s story isn’t really about AI tools. It’s about what happens when an institution decides to treat a fast-moving technology as a shared challenge rather than a compliance problem.

The pizza lunches matter as much as the provost’s committee. The faculty workshops matter as much as the enterprise software agreements. And the students who are already using AI — with or without institutional blessing — matter most of all.

“Responsible AI isn’t a technology. It’s a practice — and those practices are informed by the context in which people live.” — Michael Colaresi

That’s a sentence worth printing out and taping to the wall of every AI strategy meeting happening right now.

The tools are available. The question is whether the culture around them is being built with the same intention.