MedTech organisations risk creating more problems than they solve if they introduce artificial intelligence into fragmented data and inconsistent processes, according to discussions at this year's Veeva MedTech Summit in Amsterdam. Speakers and panellists argued that successful digital transformation depends less on how quickly AI is adopted than on the reliability of the foundations beneath it: trusted data, standardised processes, proportionate governance and clearly assigned oversight.
AI is already being applied in the sector to streamline regulatory work, improve access to information and support more efficient operations. But the consistent theme across presentations, panels and interviews was that the technology's value is constrained by the environment it runs in.
Medtronic: 'We had to earn AI'
Becki Brasher of Medtronic described a four-year programme to prepare the company for AI implementation that began with the organisation rather than the technology. The work involved harmonising processes, improving data consistency, consolidating systems and establishing governance structures. "We didn't start with AI. We had to earn AI," Brasher said during her session.
Her account illustrated why deployment into fragmented or inconsistent environments can produce outputs that are difficult to trust, validate or scale. She also stressed that technology alone cannot deliver transformation, and that employees need the knowledge, confidence and appropriate oversight to incorporate new tools into their everyday responsibilities.
Governance built in, not bolted on
The closing keynote from Erik Orzechowski and Benjamin Rochette focused on balancing faster adoption with the quality, safety and regulatory standards the industry is expected to meet. Their argument was that governance should be embedded in technology development from the outset rather than treated as a final compliance exercise.
That requires closer collaboration between IT, Quality and Regulatory functions: IT teams understand the technology and its implementation, while quality and regulatory professionals assess risks and determine the evidence needed to demonstrate compliance.
The speakers also called for proportionate governance. An AI application supporting relatively low-risk administrative activities does not necessarily need the same level of control as one influencing patient safety or regulatory decisions. Establishing responsibilities, risk assessments and controls early, they argued, allows a more flexible approach that supports innovation without overlooking obligations — framing compliance and innovation as complementary rather than opposing priorities.
Human judgement and expertise
Retaining human judgement was another recurring theme. Brasher said she expects human review to remain in place "until we can build up some trust that what we're getting, what the agents are delivering is accurate". AI can analyse large volumes of information, generate documentation and provide recommendations, but a convincing answer is not necessarily a correct one, particularly in a highly regulated environment.
During the closing keynote, the speakers compared relying on AI without understanding its underlying subject matter to using a calculator without knowing the mathematics behind it. One concern raised was the next generation of regulatory and quality professionals: if employees become accustomed to leaning on AI before developing the necessary underlying knowledge, organisations could end up with fewer people capable of challenging inaccurate or inappropriate recommendations.
Measuring value through controlled pilots
With potential applications proliferating, organisations were urged to resist pursuing every opportunity at once and instead target existing operational challenges with clear measures of success. The keynote cited time savings, reduced process cycle times, less rework, improved quality and employee adoption as ways to judge value. Controlled pilots were presented as a means of learning where AI is most effective, retaining appropriate human oversight and building confidence before expanding successful applications.
Because the technology and regulatory expectations will keep evolving, governance structures need to adapt alongside new capabilities. Success, the summit's message suggested, will depend less on the number of initiatives an organisation launches than on its ability to turn promising applications into reliable, scalable business capabilities. Orzechowski's phrase for the approach was the need to "slow down to speed up", and the event closed with a call for closer collaboration among manufacturers, technology providers, regulators and other industry stakeholders.