AI in drug development is only as reliable as the evidence behind it

The intersection of AI with medication decisions

Medication decisions represent the most interconnected workflows in healthcare, emphasizing the need for shared, consistent evidence. They are also where AI’s impact is most immediate and most fragile:

• Medication-related harm impacts 1 out of 30 patients, and is related to half of all avoidable harm in healthcare.
• 41% of physicians and 37% of nurses report using AI to check diagnoses or drug interactions.
• Healthcare industry leaders note that the industry struggles with AI output reliability for nuanced scenarios like edge cases, rare diseases, and atypical dosing.

For life sciences organizations, medication decisions have to prioritize safety considerations while also moving at speed to address growing volume and need: The global medicine market is forecasted to grow at a 5-8% CAGR through 2030, while aging populations are increasing medication complexity and polypharmacy management needs.

Patient safety and innovation are not opposing goals. The opportunity is to help life sciences teams explore emerging evidence and new possibilities while remaining clear about what is known, what is uncertain, and where expert review is needed. This creates space for responsible progress without losing sight of the hope that medical innovation can represent for patients.

In this fast-moving environment, AI often seems like a miracle solution – accelerating processes and delivering instant answers. But within healthcare, AI isn’t a substitute for medical expertise. Life sciences teams – spanning R&D, medical, access, commercial, and lifecycle – need AI-enabled workflows that are developed, not to replace clinical judgment, but to serve as a powerful asset that thrives when directly guided by it.

Patient safety in the age of AI: What drug developers need to know

AI has moved beyond experimentation, with adoption expanding across professionals and teams. However, this use often occurs at the individual or departmental level, with AI tools being applied as fragmented point solutions rather than enterprise-wide, governed approaches.

As AI expands across R&D and commercialization, life sciences organizations need to be confident that the scientific information being applied preserves its source, limitations, clinical context, and intended meaning. Trust in an AI-enabled insight should come from the integrity and transparency of the evidence behind it, not simply from how quickly or convincingly the output is presented.

Provenance and validation need to be the priorities in any AI tool selection — understanding where the information is coming from, what content it’s built on, and who is standing behind it:

• 91% of providers say it’s important that AI systems and sources for generated clinical content are validated by a human expert.
• 89% of health consumers agree.

This is especially critical for professionals working in medication-related fields and drug development, as the smallest dosing error or a convincing hallucination can impact outcomes and confidence in AI-powered drug insights.

“It’s not an accident that a key part of the Hippocratic Oath is ‘Do no harm’. Not doing harm should be top of mind as we evaluate and adopt new technologies,” says Kristen Eckler, MD, FACOG, Director of Clinical Content for UpToDate®, Wolters Kluwer Health

Keeping evidence and human expertise in AI-driven drug development

AI can support access to and interpretation of scientific information, but relevant expertise, evidence validation, transparent sourcing, governance, and accountability remain essential to responsible use.

Governed and expert-reviewed AI also benefits life sciences organizations by connecting teams – from early drug development to post-launch communications – with consistent information that helps enable more aligned decisions. This helps promote quality outcomes that support enhanced consumer safety and experience.

Evidence-grounded, reviewable AI tools can help teams access clinical interpretations with greater transparency while preserving the connection to supporting evidence:

• UpToDate® Expert AI: Provides conversational, AI-generated answers grounded in expert-authored, peer-reviewed UpToDate content, with transparent access to supporting sources, clinical assumptions, and rationale to support professional review.
• Medi-Span® Expert AI: Grounds AI-assisted medication workflows in evidence-based intelligence by applying one maintained medication foundation across two paths: AI-ready intelligence for the applications and agents organizations manage, and defined medication workflow agents that Wolters Kluwer provides and supports.

When it comes to medication dosing, generic LLMs have shown limitations in performing calculations, and even small dosing errors can create meaningful patient safety concerns. Medication-related decisions need to be shared among teams and built on governed and structured medication intelligence solutions within which drug, dosing, and clinical information are expertly reviewed. This supports life sciences organizations in prioritizing patient-safety initiatives consistently and maintaining accountability when AI is used across R&D, medical, access, commercial, and lifecycle functions.