Artificial intelligence is moving rapidly from pilot projects into day-to-day laboratory workflows. But as AI systems learn, adapt, and influence decisions that may ultimately affect end users, including patients, regulators are sharpening their focus on how these tools are validated, monitored, and governed over time.
Drawing on her experience working with laboratories at different stages of digital maturity, Kimberly Remillard, Senior Regulatory Affairs Manager at Thermo Fisher Scientific, explains how regulatory expectations are evolving – and what scientists and laboratory leaders must consider to build trust in AI-enabled and automated systems without sacrificing data integrity, accountability, or scientific rigor.
From your perspective, what are the main forces shaping how regulatory expectations for analytical and QC laboratories are evolving today?
First and foremost, the landscape is modernizing. More labs are moving from paper to digital tools like laboratory information management systems (LIMS) and electronic lab notebooks (ELN), which help them track samples, tests, and results from end to end. As a result, the automation of processes, IoT-enabled capabilities, and the seamless connectivity of digital tools and laboratory equipment through integrated platforms are becoming increasingly necessary in modern laboratories. This digitalization means that labs can minimize human error, move faster, and get higher quality results, which also helps regulators achieve their goal of getting safe and effective pharmaceuticals to market as quickly as possible. The traceability, data integrity, and efficiency afforded by digital tools are the primary drivers of regulatory expectations.
Regulators have been working to update standards and requirements to allow for digital work. Today, they’re actively encouraging the use of digital tools because they've been able to see how the benefits help them achieve their core aim: protect patients and keep product quality high.
Regulators are also looking at the rise of next-generation technologies, such as artificial intelligence (AI) and machine learning (ML), and are beginning to provide guidance into how these tools may be used, how labs can ensure quality when using them, and what role scientists play in these workflows. While there are already considerations set forth by the U.S. Food and Drug Administration (FDA) and the European Union AI Act, we know that regulators are working on additional guidance for high-risk use cases, like products that reach patients.
What aspects of day-to-day laboratory operations are being reshaped most noticeably by these shifts?
In an automated digital lab, you can see the change in daily work the moment you walk in. Filing, data entry and even things like pipetting can be digitalized and automated, significantly reducing the amount of manual work for scientists at the bench. This frees them up to focus on quality over quantity experimentation – and drive breakthroughs.
Take automated data capture, for example. Scientists can scan a barcode on their labeled sample, which minimizes the risk of translation errors, and then enter results directly into integrated LIMS, ELNs or other software. The system can flag spec results or spot changes in trends, which may indicate that there’s an issue in the experiment, as they appear. This helps teams act faster and stop small issues from turning into batch loss or delays. Scientists can even tell the system to send automated notifications to their colleagues when they’re done using the equipment, so that the lab is running at peak efficiency.
Data integrity has been a central theme for decades, yet expectations continue to evolve. What feels different about the data integrity challenges labs face today?
Data integrity has always been a priority, but today, regulators require higher standards of evidence. Inspectors now look for proof of data integrity at each step, not just in the final report.
When the U.S. FDA’s 21 CFR Part 11 regulations established the criteria for electronic data to be as reliable as paper records in the late 1990s, most technology on the market couldn't meet the needs. But that’s not the case today. Regulators now expect labs to meet Attributable, Legible, Contemporaneous, Original and Accurate Plus Plus (ALCOA++) standards across the full data life cycle, which means that the process – from the first test to long term storage – needs to be traceable and documented.
In well-designed digital laboratory environments, many compliance controls can be embedded directly into workflows. Features, such as electronic signatures and established access rights, give a clear view of where data lives and who can view or change it. This helps labs ensure data integrity in all that they do
As labs begin exploring AI-enabled tools, what new regulatory questions or uncertainties are emerging?
AI-enabled tools bring clear gains, yet they also raise new questions. Validation is a key consideration, as labs must show that an AI model is accurate, reliable, and fit for its stated use. And they must show this not just once, but over time as the model learns or as data shifts.
In high-risk environments, monitoring the inputs that AI uses are incredibly important. Labs must explain what input the AI tools use, what logic it follows, who trained it, and where its limits are, including where human-in-the-loop steps are required. This is critical for AI-enabled technologies that play a role in developing products for patients.
Regulatory guidance around AI is still evolving, as are the AI-enabled technologies. However, most have a risk-based credibility assessment framework that labs can use to establish or evaluate an AI model for specific use cases. For labs, regulators have stressed the need for human-in-the-loop workflows, which means that a trained person must still review and sign off on any AI-driven outputs to ensure integrity and compliance.
How realistic is the idea of “compliance by design” in today’s laboratories?
Compliance by design means using tools that have built-in features to help scientists track workflows and avoid missteps – modern LIMS and other digital tools already have these features. So, yes, I do see a real path to compliance by design. In fact, many labs are already on that path, but every lab is different. For those who are just starting to move toward building their automated digital labs, configurable software, tailored to their specific workflows, can help ensure that compliance is baked into their processes.
For compliance by design to work effectively, labs must invest in modern digital tools, configure them to match their process, validate that they’re fit for purpose, and have a way to regularly maintain any updates. That’s why the choice of digital systems – and how they are implemented and maintained – is so important. Well-designed platforms can support regulatory updates, audits, and upgrades, provided they are properly configured, validated, and governed over time.
How do you see emerging frameworks – such as the EU AI Act or the FDA’s Computer Software Assurance (CSA) initiative – shaping expectations for analytical and QC labs, especially those using complex software environments?
While many regulations are still in flux, these new frameworks will set the tone for how labs will run complex software and AI for years to come. For example, the EU AI Act categorizes most AI applications within labs as high-risk, which means that they are subject to strict requirements. Under this act, labs must establish a risk management system, conduct data governance training and validation, draw up technical documentation to demonstrate compliance, design their system for automatic recordkeeping, provide instructions for downstream use, implement human oversight, establish a quality management system, and achieve appropriate levels of accuracy, robustness and cybersecurity. This goes back to compliance by design – labs must choose tools that can help them meet these requirements.
The FDA’s Computer Software Assurance (CSA) initiative compliments this framework. It pushes labs to use a risk-based approach to establish and maintain the reliability and safety of their software throughout its lifecycle. The aim is to have quality testing rather than quantity testing. To reduce the testing effort, several steps must be taken to ensure the successful implementation of the CSA approach, such as identifying the intended use, risk determination, and proper documentation of assurance activities.
As compliance becomes increasingly intertwined with digital tools and data governance, how are the skills and responsibilities expected of analytical scientists and lab managers changing?
With good tools, scientists and lab managers can focus more of their time on what matters most – science. While there’s the time needed upfront to adjust to new tools and software systems, the benefits vastly outweigh the initial input. Across the lab, scientists will be able to eliminate many of their time-consuming manual tasks and, instead, focus on experimentation and analyzing their data. As more technology becomes available in a digital format, there will be even more opportunities for them to focus on science and quality in their day-to-day.
To get started, labs should invest in clear, hands-on training plans and build open lines of communication between scientists, information technology groups, and quality experts to support onboarding, upskilling, and change management.
The goal is to build an automated digital lab that enables scientific orchestration. A holistically connected lab will help scientists overcome capability gaps, ensure data readiness, stay in lockstep with new rules and drive strong, high-quality science that accelerates breakthrough discoveries.
What should labs be prioritizing as they prepare for the future?
Regardless of where labs are in their digital journey, there are a few core areas to keep in mind to ensure long term success. First, labs should invest in digital systems that can help them move away from paper-based work. These systems should be connected and compliance-ready so that the lab can keep a full trace from start to end. Managers should build strong partnerships with trusted technology suppliers who can help with audits, upgrades, and ensure that the lab stays up to date on compliance and security as guidelines evolve. Lastly, look to the future by planning for AI use and automation. By building a strong digital foundation, labs will be able to scale operations and integrate new technologies when needed.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
