AI Is Quietly Reshaping Puerto Rico's Pharma Industry — Regulators Want Humans Kept in the Loop

Puerto Rico's economy has a well-known pillar, and it isn't tourism. Pharmaceutical manufacturing accounted for $48.3 billion — 74% of the island's total exports — in fiscal 2024, and the sector contributes roughly 30% of Puerto Rico's GDP, putting the island second only to Indiana among all U.S. states and territories in pharmaceutical export value. So when the Pharmaceutical Industry Association held its 24th regulatory conference this week under the theme "Advancing compliance and operational excellence," the conversation about artificial intelligence happening inside those sessions wasn't a side topic. It was a discussion about the technology reshaping the industry that anchors the island's economy.

An Industry Built on Precision, Meeting a Technology Built on Probability

Pharmaceutical manufacturing runs on documentation, deviation tracking, and investigation trails — the unglamorous infrastructure that keeps a drug batch traceable and a regulatory inspection survivable. That's exactly the kind of work generative and analytical AI is good at accelerating: sorting through large volumes of data, flagging deviations, spotting trends across production runs, and supporting the investigations that follow when something doesn't match spec. PIA's 11 member companies operate 13 manufacturing plants across seven Puerto Rico municipalities, and the throughline across the conference was that AI is already doing real work inside those facilities, not sitting in a pilot-program purgatory.

But pharma is also an industry where a wrong answer isn't a bad recommendation — it can be a contaminated batch, a mislabeled dosage, or a failed inspection with regulatory consequences. That tension between AI's appetite for speed and pharma's zero-tolerance relationship with error was the spine of the conference's AI discussion, distilled into one line that came up repeatedly: "AI can speed the work while people retain responsibility for the decisions."

The FDA's Cautionary Tale

It wasn't all industry optimism. Marian E. Ramírez of the FDA's Office of Inspections and Investigations presented a case study built specifically to illustrate what goes wrong when AI operates without professional supervision or adequate controls — a pointed inclusion at a conference themed around compliance. The specifics of the case weren't the point so much as the message: a regulator's presence at an industry conference, delivering a scenario about AI failure rather than AI promise, is itself a signal about where oversight bodies think the risk actually sits. It's not that AI can't help with deviation analysis or trend detection — the industry consensus in the room agreed it can — it's that deploying it without validation protocols, established controls, and mandatory human review is how a helpful tool turns into a compliance incident.

That pairing — an industry association demonstrating adoption while a federal regulator demonstrates the failure mode — is a fairly rare thing to see on the same conference agenda, and it suggests Puerto Rico's pharma sector is trying to get ahead of an oversight conversation rather than wait for it to arrive as an enforcement action.

What "Responsible Adoption" Looks Like in Practice

The requirements that came out of the conference weren't abstract. Validation protocols, established controls, and mandatory human review processes were named specifically as the conditions under which AI tools are acceptable in this environment — a more concrete framework than the vague "responsible AI" language that shows up in a lot of corporate messaging. For an industry this heavily regulated, that specificity matters: an FDA inspector isn't going to accept "we used AI to speed up the review" as an answer on its own. They're going to ask what validated the model's output, who reviewed it, and what would have caught it if the model was wrong.

David Thompson, a PIA leader who spoke at the conference, framed the broader posture as "operational excellence is not a destination, but a continuous journey" — a line that could read as conference boilerplate, except it lands differently next to an FDA official's failure case study on the same agenda. The subtext is that AI adoption in this sector isn't a one-time technology rollout to check off; it's an ongoing governance problem that has to be maintained at the same rigor as any other compliance process.

Why This Matters Beyond Pharma

Puerto Rico's pharmaceutical sector has spent decades building a reputation for regulatory rigor — it's part of why global pharma companies manufacture here rather than in less scrutinized jurisdictions. How that sector handles AI adoption is a preview of a broader question every regulated industry on the island, and frankly every regulated industry anywhere, is going to have to answer: how do you get the productivity benefits of AI without quietly eroding the human accountability that regulatory trust is built on? Puerto Rico's pharma manufacturers, representing nearly three-quarters of the island's export economy, don't have the luxury of getting that answer wrong. The conference suggests they know it.

Takeaway

The story out of PIA's conference isn't "pharma is using AI" — that part was already underway. It's that Puerto Rico's most economically important industry is trying to build AI adoption and regulatory accountability at the same time, with an FDA official in the room to make sure the second part doesn't get skipped. For an island economy this dependent on one heavily regulated sector, getting that balance right isn't just good practice — it's close to existential.