US Life Science Tools & Diagnostics: How will pharma’s use of AI in R&D affect CROs + preclinical tools? (expert webinar takeaways)

Research evidence excerpt

US Life Science Tools & Diagnostics: How will pharma’s use of AI in R&D affect CROs + preclinical tools? (expert webinar takeaways)

Eve Burstein +1 917 344 8313 eve.burstein@bernsteinsg.com 17 July 2026

4. CROs are keeping more of the upside margin now, but this is likely not

sustainable. CROs may initially retain a meaningful portion of AI-generated savings

because they are absorbing the investment and implementation risk (likely ~60-70% of

upside margin today). As capabilities become standardized and competition increases,

sponsors will likely capture more of those savings through lower pricing.

Takeaways on preclinical research:

A. AI may generate more ideas, but it does not generate evidence. Every proposed

drug candidate still requires evidence including efficacy testing, safety testing, bioanalytical

work, toxicology studies, and ultimately clinical validation. The expert repeatedly returned

to the distinction between generating hypotheses and generating evidence; AI may become

increasingly effective at the former, but regulatory approval remains dependent on the

latter.

B. More productive R&D could create a volume tailwind. If development becomes

cheaper and more programs become economically viable, an increase in number of assets

being developed could potentially offset any reduced spend on development per asset.

C. The more immediate productivity gains are likely to come from the administrative

work of science rather than the laboratory work. Unlike some of the more aggressive

AI narratives currently circulating, the discussion pointed toward report generation,

documentation, regulatory preparation, meeting administration, project coordination, and

data processing as the most obvious sources of efficiency. For example, with AI scientists