From data to deals: AI’s impact on life sciences investing

There was a noticeable shift in tone compared with a similar discussion at our symposium held just 12 months ago.

Previous audience polling at McDermott events indicated some hesitation about the future impact of AI.

But at this year’s session, prior hesitation seemed to have vanished – around 40% of attendees described AI as already embedded throughout their organisations.

This acceleration, and whether regulation has a realistic chance of keeping up with the pace of change, became one of the defining themes of the panel.

Historically, healthcare organisations have not adopted technology quickly in the past, even when the case for doing so was obvious. Indeed, the healthcare industry has often been good at developing products and technologies, but less good at real adoption. Based on our discussion, AI feels like a very different case.

To take one example raised in our panel: electronic medical records took seven years to achieve 50% penetration across US healthcare systems, despite regulatory pressure and government incentives. AI scribes reached the same level in roughly 18 months. The explanation is not complicated. Most clinicians immediately understood the value proposition because it addressed something painfully familiar: administrative overload.

The panel recognized that a key driver for this adoption is the operational pressure facing businesses, and the need to find efficiencies.

Against that backdrop, AI can serve as a practical tool for reducing friction in systems that already feel overstretched.

Whilst some discussions tend to focus on most futuristic use cases, the panel concentrated on the fact that the technologies with most traction are often those that speed up “behind the scenes” support, such as workflow automation, scheduling, patient communication, documentation, data extraction and operational support.

One distinguishing feature in the fast adoption is that usage appears to be driven from the ground up. One panellist observed that, depending on the jurisdiction, somewhere between 25% and 75% of healthcare practitioners may already be using these tools to help understand patient histories, clinical questions or treatment pathways.

Clinicians are using these systems because they find them useful, not because healthcare institutions have necessarily completed lengthy procurements or usage validation exercises. In many ways, that is what distinguishes this moment from previous waves of healthcare technology adoption, and should be a key focus for boards.

At the same time, the panel noted that many aspects of healthcare and life sciences companies remain inefficient and that there are challenges with the large volumes of unstructured data For all the excitement around the opportunities of AI, including from drug discovery to clinical trial efficiencies, the reality is that huge volumes of healthcare data remain effectively unusable. One panellist observed that around 80% of critical medical data remains unstructured, sitting in PDFs, free text and fragmented hospital systems.

This gap between ambition and infrastructure may create opportunities for companies focused on structuring, cleaning and contextualising healthcare data so it can be used effectively by AI systems downstream.

The discussion also looked at investment strategy and defensibility, given AI developments in AI including agentic AI.

The last few years has seen significant investment in single-purpose radiology models. Now, with the potential of large foundation models, it seems feasible that AI tools may develop models that perform across multiple pathology areas.

This has obvious implications for investors. The challenge is no longer as simple as proving that a model works. Increasingly, companies need to demonstrate how and why they will remain relevant in an environment where underlying capabilities are evolving so quickly.

Ultimately, differentiated datasets, integration into healthcare systems and the ability to build reliable, auditable AI ecosystems may matter more than standalone point solutions.

The panel predicted a shift from isolated models toward ecosystems of agentic systems where future differentiation depends on how effectively businesses orchestrate within complex healthcare environments.

The panel also recognised that AI requires meaningful overside and regulation but had concerns that law may have a dampening effect on innovation. There was concern that the pace of legislation enactment lags behind technological change so that law is out of date before it is in force.

The EU AI Act, which is not yet fully in force, received some scrutiny. Questions were raised about the risk of trying to regulate technologies that are already evolving beyond the categories legislators originally had in mind, not least because the EU AI Act imposes a double layer of regulation on software (including AI) as a medical device. This additional layer of cost and certification might provide comfort to users and legislators but risks deterring innovators and delaying efficiencies in healthcare systems.

A key takeaway is that AI is no longer at the edge of healthcare strategy. The challenge is not persuading the market that AI matters. It is determining where sustainable value will sit in an ecosystem evolving so quickly that many of the assumptions underpinning today’s technology, regulation and business models may look very different again within a remarkably short period of time.