EP 15 - AI Can Support Clinical Research, But It Cannot Replace Clinical Thinking
AI Can Support Clinical Research, But It Cannot Replace Clinical Thinking
It was a great pleasure speaking with Andreas Beust, Andi, CEO of GCP Service, about what is happening in the CRO world, clinical research, AI, and medical devices. One point stayed with me strongly after our conversation: clinical research is changing fast, but not every faster solution is automatically a better one.
Every week, there is a new AI tool or digital health application promising to make clinical trials faster and easier. Technology can help with document creation, language quality, and reducing mistakes. But in clinical research, faster is not always better.
Andi raised this concern directly. When companies use AI to prepare study documents, the problem is often not the first draft itself. The problem is that without experience from hundreds of trials, you may not know where the document is wrong, where the gaps are, or where you need to think again.
This is the real risk for smaller medical device and biotech companies. A protocol can look professional and still miss the real clinical or regulatory question. An informed consent form can be well-written yet still not be appropriate for the patient group. A statistical plan can appear complete and still fail to support the study objective.
For medical devices, Andi made another important point: everything comes back to the labeling claim. Choosing the right claim can mean a smaller study, a shorter timeline, and faster market access. But if the study does not support the claim, the company may collect data that cannot be used. As Andi said clearly, if a company runs a trial it cannot use later, this can very often be the end of the company.
Many smaller companies depend on external funding, which has become much harder to secure. Inflation, higher interest rates, and global uncertainty have made investors more careful. In this environment, the temptation to use AI as a quick solution is understandable. But AI does not have real-world experience. It does not sit with a sponsor and ask whether the study truly supports the future labeling claim.
Clinical thinking means having people who understand the patient, the endpoint, the regulator, and the investor. People who can say early enough: "This looks good on paper, but it will not work in reality." Europe adds another layer of complexity. It remains attractive for international trials due to quality and cost, but navigating different languages and local requirements across Germany, France, Spain, Poland, and other countries demands real preparation.
The same challenge exists in digital health. There are countless mental health apps, lifestyle tools, and software products that function as medical devices. Sometimes, even their creators cannot clearly explain what makes them different or why they should exist.
AI will help clinical research. It will simplify some processes and reduce repetitive work. But it should not replace the people who know how trials really work. The future should not be AI alone. It should be AI together with clinical research experience.
Thank you, Andi, for sharing your practical perspective. Evidence must stay at the center. Without it, we are not building innovation. We are only building nice documents.
00:31 CRO Market Consolidation
02:39 AI Hype In Clinical Research
03:37 Where AI Helps Today
06:36 Risks Of AI Written Protocols
09:39 Medical Devices And SaMD Boom
11:12 Funding Uncertainty Outlook
13:40 Investing In Digital Health
17:25 App Overload And Differentiation
21:03 How They Pick Investments
23:13 GCP Service End To End CRO
25:50 Why Europe Still Attracts US Trials
28:02 Europe Complexity Languages
30:34 What Makes Trials Succeed
33:39 Clinical Trials In 10 Years
35:39 Personal Vision And Wrap Up
Links:
Pavlina Walter: https://www.linkedin.com/in/pavlinawalter/
Website: PavlinaWalter.com [http://pavlinawalter.com]
Guest:
CEO Dr. Andreas Beust: https://www.linkedin.com/in/andreas-beust-8bb32295/ [https://www.linkedin.com/in/andreas-beust-8bb32295/]