AI in Regulatory Affairs: What our Panel Conversations Are Actually Revealing

By Samantha Dowse, Managing Partner EMEA

The Pharma Tech Summit 2026 brought together a broad cross-section of the industry: academics, media, and a significant number of AI technology vendors.

Our expert team contributed to the panel on Regulatory Guidance Using AI, and what struck us most was the specificity of the concerns being raised. The conversation has moved on from "should we use AI?" to harder, more practical questions about where it works, where it doesn't, and what responsible adoption looks like in a regulatory context.

Here are the themes that cut through.

AI is changing design, not approval

The clearest area of genuine traction right now is early-stage design and predictive modelling. For CDMOs, AI tools are delivering reductions in experimental workload i.e. helping teams model outcomes, identify risks earlier, and make better decisions about where to focus resource before work reaches the bench. This is reflected in Developability and Manufacturability packages that are being conducted to streamline the analytics physically performed.

There's an important line being drawn by regulators, and it came up repeatedly during the panel: AI-generated outputs are not, at this stage, being accepted at later development stages without robust, human-executed data to support them. The evidence is still required, such as validation and stability studies. What AI changes is how efficiently and intelligently you build toward it — not whether you need it.

For organisations mapping their AI strategy in development, this distinction matters. Tools that help you design smarter are valuable. Tools that promise to reduce your evidential burden are, for now, overpromising.

Human regulatory expertise isn't being replaced everywhere

One of the more nuanced points raised during the panel was about the role of the regulatory professional in an AI-assisted workflow. There's a version of the AI adoption story that frames human review as a bottleneck; a legacy step that will eventually be automated away. That's not what the evidence supports, and it's not what experienced practitioners are seeing on the ground.

The firms with the most confidence are treating qualified regulatory professionals as the layer that gives AI outputs their credibility. Every AI-produced submission, dataset, or analysis still needs expert review.

With these changes in processes, experienced regulatory professionals who understand how to work with AI tools, interrogate their outputs, and apply appropriate judgement are becoming more valuable, not less. The question for organisations is whether they're investing in that capability alongside their technology.

The AI conversation has quietly expanded beyond R&D for Biotech

Something that perhaps doesn't get enough attention in the biopharma AI discussion: adoption is happening with considerable pace in functions well outside the laboratory.

Finance and business development teams are implementing AI tools, often with less fanfare and (in some cases) more operational agility than R&D functions, where the regulatory stakes naturally create more caution. The organisations we saw navigating this well aren't waiting for a single top-down AI strategy to land before they act. They're enabling function-level adoption, with governance frameworks that travel with each use case rather than sitting at the centre, maintaining the “Human in the Loop” role as detailed by the EMA.

This matters because it shifts the framing from functional to organisational. AI in pharma isn't only a scientific question or a regulatory question, instead it’s a strategic business question. How decisions get made, how risk gets owned, and how different functions share what they're learning from early adoption will shape outcomes as much as the technology itself.

What this means in practice

The hype around AI hasn't gone away, in fact, the most useful discussions are happening in the space between the promise and the current capabilities. Practitioners are working through specific problems and building practical knowledge about what works.

GreyRigge’s free counsel:

• Start where the evidence is strongest. Early-stage product design, predictive modelling, manufacturability and operational efficiency are the areas with the most mature use cases. That's where investment is most likely to return value in the near term.
• Build the human layer intentionally. AI adoption without qualified oversight can increase regulatory risk, without employees knowing about it. AI redistributes data in ways that can be harder to manage. Investing in regulatory expertise that can work effectively alongside AI tools is as important as the technology investment itself, “Human in the Loop”.
• Don't wait for a unified strategy to begin. Function-level adoption with appropriate governance is how many organisations are making progress. Waiting for perfect conditions tends to mean falling behind.
• Watch the late-stage acceptance question closely. Regulatory openness to AI outputs at later development stages will evolve, but it will do so incrementally and with significant scrutiny. Organisations building toward that future need to be engaged with the regulatory dialogue, not just the technology.

These are questions our team works through with clients across the US, UK, Europe, and Japan regularly. If any of this describes challenges you're navigating, we'd be glad to talk and support your project.

Contact Us