One of the goals most trumpeted for AI is the curing of human diseases. Let’s look ahead and see both what might encourage and challenge us as we pursue this goal. As practical AI applications in the life sciences industry and in medicine begin to proliferate, we see that, in principle, AI can bring greater productivity and faster progress to both.
But as we dig deeper, we see that the nature of AI’s impact is rather different across the two healthcare sectors, and therefore the barriers to adoption are too.
In life sciences—taking pharmaceuticals as a prime example—AI can, and I believe will, result in the discovery of more effective new drugs and vaccines, accelerating their journey to potential approval. Since new products, and their profitability whilst they remain on patent, are the financial lifeblood of the industry, that should spell an era of ‘onward and upward’ progress. So in that sense the impact of AI is fully aligned with the business model of the industry. But the progress that AI drives will inevitably meet systemic antibodies.
First, the faster flow of potential new products will severely test the capacity of healthcare systems to conduct the necessary clinical trials. In all diseases, there is a finite number of patients that qualify for such trials, and already there is stiff competition to find and enrol them. That may prove to be a significant bottleneck.
Next comes regulatory assessment and approval. The regulatory system is already strained to meet the current flow of products; will there be the political will to expand the number of assessors and advisory committee members? And, crucially, when this larger flow of effective new products reaches the market, will payors be happy to reimburse them?
So more, better and faster life science innovations lie ahead, provided the system to review, approve and reimburse them can adjust to the flow.
When we turn to AI’s adoption by the health systems themselves, I see quite different potential barriers, and therefore more substantial issues for leaders to resolve.
Most fundamentally, can medical education keep pace with AI’s advance—and train the next generation of doctors on how best to meld human and machine intelligence? And, given that AI tools have the capacity to enhance the productivity of clinicians, will they be seen that way, or as a threat to medics’ job security? Will medical societies and those negotiating for clinicians welcome this trend or oppose it?
Furthermore, AI will clash directly with existing healthcare business models. AI will help identify diseases earlier and often prevent them, whilst most hospitals are currently financially rewarded for carrying out procedures to deal with late-stage disease. Seen in that light, AI could represent an overall financial debit rather than a credit for hospital managers, creating a structural friction against adoption of those AI tools that provide early detection and timely intervention.
Finally, to fulfil their potential, AI tools will force changes in operational workflows for which clinicians and managers may simply not be prepared. We already see this barrier to adoption of non-AI technologies that call for new pathways of care and new skills from clinicians. And accountability for error will be another point of potential friction. Just as in legal disputes around collisions involving self-driving cars, there will inevitably be high-profile incidents blamed on AI that could cause a backlash in terms of regulation or public trust.
Leaders in both life sciences and medical care need to be clear-minded about these barriers and risks, and not simply encourage their staff to adopt AI willy-nilly. The future is exciting, but experience should teach us not to take its arrival and productive adoption for granted.
Which of these bottlenecks—regulatory strain, clinical trial capacity, or hospital financial models—do you believe will pose the greatest challenge to AI adoption in healthcare? I'd love to hear your thoughts in the comments.
