Last year, my department started rethinking professional development at CUNY Libraries. Instead of one-off trainings, we began building programs that grow shared technical capacity across campuses.

This semester, that work has taken a hands-on turn.

We’re three weeks into a 16-week peer mentoring cohort focused on agentic AI and what it actually means for library workflows.

The cohort is sponsored by the Office of Library Services. My team typically works on infrastructure—systems, integrations, APIs—but infrastructure isn’t only technical. It’s human. If vendors are going to build AI into library systems, we need librarians who understand it well enough to evaluate it, question it, and, when necessary, build alternatives.

The idea builds on our Alma Extensibility Task Force, where librarians use Alma’s REST APIs to solve systemwide problems. That group has quietly produced automations and integrations Alma doesn’t natively support. The AI cohort expands that model beyond Alma and beyond the people who already see themselves as developers.

Thirteen faculty and staff from nine CUNY campuses are spending the semester building AI-enabled tools tied to real needs: automating publication tracking for repositories, creating accessibility checks for course content, enhancing discovery, streamlining reserves, improving shared data access.

Skill levels range from beginners to experienced coders. That mix is intentional. Peer learning works best when people bring different strengths to the table.

Several projects I’ve wanted to tackle for years—like externalizing Alma letters into version-controlled infrastructure—are finally moving. The difference isn’t just better tools. It’s structured time and a cohort of colleagues building alongside each other.

My role is mostly scaffolding: coordinating logistics, maintaining the Teams space, making sure experimentation feels possible but not overwhelming. It’s administrative, sure, but it’s also strategic. Capacity building requires design.

Already, the Teams channel is active with troubleshooting, API questions, and real workflow conversations. People are making time for this despite heavy workloads. That matters.

The goal isn’t just AI literacy. It’s distributed expertise.

In a 26-campus system, we need librarians who can assess vendor claims, understand technical tradeoffs, and decide when to build instead of buy. That kind of fluency is infrastructure.

We’re only three weeks in. But it’s clear this isn’t about chasing a trend.

It’s about expanding who gets to build—and making sure librarians remain active participants in shaping the systems that shape us.

More soon.