TLDR
- Internal and external pressures are rapidly converging on the life sciences and its CMC leaders. As other healthcare verticals normalize AI, pharma leaders face growing questions about how they’re keeping up — and why many are not.
- For many, regulatory uncertainty is still the greatest restraint. The industry wants speed, but FDA turmoil and questions around acceptable AI use cases, governance, and review expectations are all sapping leaders’ confidence to “move fast.”
- The hype-to-implementation shift continues to raise the compliance bar. Once AI touches CMC workflows, it’s about more than efficiency and speed — it’s about traceability, validation, monitoring, and explaining outputs under scrutiny.
- The bottom line for CMC: Now more than ever, structured data and knowledge are the prerequisite for “AI-ready.” If your data still lives in fragmented systems, static formats, and tribal memory, you’re not ready to govern, validate, or confidently scale AI.
Well, that was… different?
Like most JPM-watchers, we had the popcorn out for some big headlines — a true mega-merger, one game-changing platform launch, a funding round the size of a small national GDP. But while the credits rolled without a breakout star, there was still plenty of action in the life sciences.
The energy was definitely there, but this year, it was more nervous than usual — less “swashbuckling dealmakers,” more “seasoned submariners watching the sonar.” Between the quieter headlines, the vibe was shifting: From the excitement of dizzying AI velocity to the realities of integration, governance, and compliant validation. Everyone knows AI is here to stay — but no one’s exactly sure where it’s going next.
For drug developers and manufacturers, though, that uncertainty carries its own clear message. With AI transformation moving on its own momentum, AI readiness is no longer an aspiration: It’s a mission-critical priority that demands a plan initiated yesterday. And that spells tricky vibes for a lot of CMC programs.
The AI pressure is rising. And it’s no longer internal
One big recurring theme at JPM26: The pressure to adopt and deploy AI is coming from outside drug development organizations as much as from within.
Clinical development teams are already reaping the benefits of AI use cases in trial design, recruitment optimization, and protocol feasibility, while big investments in AI-powered drug discovery are bearing fruit, too. Meanwhile, pharma’s “cousins” in the healthcare delivery space are blowing the AI adoption curve, deploying effective solutions for documentation, triage, coding support, clinical decision support, and admin workflows as well. The net effect: A healthcare ecosystem where most domains now treat AI as normal infrastructure, not an experiment.
With one notable exception.
As the AI momentum at JPM made clear, drug development and manufacturing teams are more on the spot than ever. With so many healthcare domains racing toward an AI event horizon — and already delivering measurable gains — laggards are facing ever-louder questions from investors, boards, and shareholders. Where are we deploying AI in the lifecycle? What’s our plan to scale it, not just pilot it? And perhaps most importantly, when will we have the data maturity to do any of this responsibly?
At JPM26, it felt like the domains surrounding CMC collectively moved from debating whether AI matters to confronting a more uncomfortable reality: successful, at-scale AI adoption is becoming a reputational and operational expectation and a compliance must-have. The only debate left is how quickly drug developers and manufacturers implement it without creating new risks.
And this year, that risk had a specific new flavor.
As the AI momentum at JPM made clear, drug development and manufacturing teams are more on the spot than ever. With so many healthcare domains racing toward an AI event horizon — and already delivering measurable gains — laggards are facing ever-louder questions from investors, boards, and shareholders.
Regulatory uncertainty: Heartburn for the whole drug development and manufacturing enterprise
To race full-speed toward any destination, AI or otherwise, you need clear lanes, guardrails, and track signals. And for drug developers and manufacturers, they’re pretty hard to see right now.
JPM26 made that acutely clear: the AI urgency in drugmaking is colliding head-on with a period of heightened regulatory uncertainty, with the FDA leading the disruptive AI charge precious little guidance for the industry.
A glimmer of direction — or at least directional expectation-setting — emerged on Day 3, in the form of joint FDA-EMA “guiding principles” for AI use in drug development. But these broad strokes do little to relieve the greater pressure point: For drug developers and manufacturers, regulatory expectations for AI-enabled processes are racing ahead of regulatory clarity on compliance “safe zones” for AI use cases.
And in a compliance-driven domain like CMC, that mismatch isn’t just inconvenient — it’s existential.
JPM week did nothing to calm the chatter about turmoil in the FDA environment or the uncertainty about the consistency of engagement and review norms — both of which were amplified by the FDA’s continued headlong push into AI. For anxious drug developers and manufacturers, the message seems to be: “Adopt AI ASAP across the product lifecycle and ensure that it’s governed, explainable where needed, monitored, and supported by appropriate controls. Exact definitions of ‘governance,’ ‘explainable,’ and ‘appropriate’ TBD.”
Alkaseltzer, anyone?
Sensing those existential stakes, most drug developers and manufacturers are making any headway they can, feeling their way toward AI use cases that can survive a quality review, a model audit, and regulatory questions that start with “walk me through…” It’s a higher-risk path than any industry leader would choose, but it’s the only one open for CMC programs — and one leads them straight toward one of their most persistent challenges.
AI is forcing a reckoning: Most CMC knowledge still isn’t agent-ready
In many JPM26 conversations, there was a consistent theme that echoed in many CMC programs. The market and regulators want AI outcomes — faster decisions, fewer deviations, smarter tech transfer, improved comparability strategies, more robust control strategies — but most CMC data isn’t ready to deliver them.
For far too many of those programs, the lifeblood of AI still lives in static documents, scattered spreadsheets, and institutional memory — “ask Priya, she knows which version of that method history is the right one.” One look at the average drug developer’s OneDrive will tell you just how real those challenges still are. And if AI is unforgiving about one thing, it’s input quality.
As any LLM will happily do, AI can summarize a PDF, Word doc, or Excel matrix. Or parse medical records and health data based on accepted data standards (sort of). But it can’t magically convert inconsistent terminology, missing metadata, unclear provenance, and implicit process knowledge into a reliable basis for regulated decision-making. At best, it produces plausible answers. At worst, it produces plausible answers with confidence — arguably the most dangerous kind.
But that reality is precisely what many CMC programs are confronting. With a board demanding to know when they can call their pipeline “AI-powered.”
Yes, this JPM week, CMC leaders couldn’t escape the hard truth: The more serious AI becomes, the less tolerance there is for CMC information that can’t be traced, reconciled, or reused across programs, sites, and regulatory pathways. The pressure to mature CMC data ecosystems — to shift them from yesterday’s fragmentation to harmonized, scalable frameworks — is now beyond acute.
And there’s only one way to relieve it.
In many JPM26 conversations, there was a consistent theme that echoed in many CMC programs. The market and regulators want AI outcomes — faster decisions, fewer deviations, smarter tech transfer, improved comparability strategies, more robust control strategies — but most CMC data isn’t ready to deliver them.
With collaboration becoming the standard, data standardization is the new golden fleece
To see these data exchange challenges on fullest display, look no further than the handoff points in the product lifecycle. All 95,344,394 of them.
It’s a reality CMC leaders live every day: for better or worse, drug development and manufacturing is more deeply networked than ever. Global sites, CDMOs, external labs, raw material suppliers, device partners, and perhaps most importantly, regulators — they’re all part of the continuous data exchange that powers the drug product lifecycle. Efficient, systematic, and comprehensive handoffs are essential to the function of these networks, as well as successful AI infusions.
And every one of them is also a chance to lose meaning, break lineages, obscure context, and conflate key datapoints. In other words, to turn potential AI-powering data pipelines into snarled information garden hoses.
So it was no surprise that data standardization efforts are resurfacing with renewed urgency at JPM26 — not as an IT clean-up project, but as a collaboration and velocity mandate. For CMC leaders, the signal is clear: It’s time to move beyond “document exchange” toward shared, structured approaches to regulatory interaction and partner collaboration, like those clearly foretold by regulatory initiatives like the PRISM program.
Standardization, in other words, is no longer a philosophical prompt for industry panel discussions. In today’s highly distributed operating model — and on today’s AI adoption curve — interoperable data is already a strategic necessity.
Fortunately, initiatives like Project Artemis are already paving the road to harmonized, exchange-ready data frameworks. Establishing and achieving that standardization goal won’t just make the current networked life cycle work better: It’s essential to unlocking AI’s full potential for drug developers and manufacturers.
And once “compliant AI use” has been rigorously clarified, we’ll really be getting somewhere.
What leaders should do now: Shifting from data-driven tools to mature data infrastructure
In their own takeaways from JPM26, the experts at Galen Growth put it beautifully: Governed AI is becoming part of regulatory expectation, not a voluntary best practice. For drug developers, drug manufacturers, and their partners, AI capability will be evaluated alongside traceability, validation, monitoring, and documentation. For drug developers, drug manufacturers, and their partners, “‘We use AI’ is not a differentiator; ‘we can defend our AI under audit’ is.”
So what can CMC leaders do now to prepare for this new reality? As JPM26 made clear, the answer is not “pick the right model.” It’s designing and implementing a data maturity program that spans data, process, and governance, and enables governance, explainability, and validation. A strong starting blueprint could look like this:
1) Establish your CMC knowledge backbone
Identify the CMC objects that matter most to your organization — materials, methods, process parameters, specifications, CQAs, CPPs, deviations, change controls, stability outcomes — and formalize how they relate. And move them to a structured framework ASAP.
2) Prioritize “lineage-first” data architecture
If you can’t explain where a number came from, when it changed, and who approved it, AI will only amplify that weakness. Establish lineage as a first-class requirement, not a reporting add-on — something a Digital CMC platform like QbDVision does by default.
3) Standardize partner data exchange where it counts
Standardizing what matters most is the first step in any smart harmonization strategy. Start with high-impact exchanges: tech transfer packages, method validation history, comparability-relevant data, and key control strategy elements. As players like Bayer and Viralgen have found, this approach can deliver big benefits fast.
4) Build an AI governance narrative you can defend
Not a glossy slide, a real narrative: what models do, what they don’t do, how outputs are reviewed, what triggers revalidation, how monitoring works, and how you prevent “shadow AI” from creeping into regulated decisions. As the FDA has already made clear, they expect glass boxes, not black ones. Your governance strategy — and how it’s implemented in your selected data framework — is essential to enabling that visibility and clarity.
Done well, this doesn’t just enable AI. It strengthens the core operating model of CMC by reducing friction, improving reuse, and turning knowledge into a productivity-powering asset — not another SharePoint scavenger hunt.
A quiet JPM still carried a deafening warning
While JPM may not have delivered that single, year-defining headline, it still put a much-needed spotlight on some of the biggest hurdles for 2026. With the industry charging from AI hype to AI implementation — and regulators shining a wobbly light on the road ahead — everyone’s racing toward a new set of operational realities, compliance expectations, and data necessities. In a car where, to borrow a phrase, “who said anything about safe?”
But for CMC leaders, the biggest JPM26 takeaway is simple and urgent: The more real and practical AI becomes, the less optional structured CMC data becomes. AI readiness is no longer a future initiative: it’s table stakes for the speed, resilience, agility, and regulatory process the market now mandates. The only question is how quickly, responsibly, and effectively you can deploy it.
See you at the races!
GET IN TOUCH
Ready to get your CMC program AI-ready?
Reach out any time to learn how QbDVision can help you create the foundation of structured data and knowledge that all successful AI deployments depend on.
