The increased accessibility of generative artificial intelligence (Gen AI) models like OpenAI’s ChatGPT and DALL-E has captured the popular imagination. You’d be hard-pressed now to find any leadership team in any sector that isn’t at least considering ways to leverage Gen AI.
The technology uses patterns learned from data to create new content. Most industries have applied it to routine back office or administrative tasks, such as note-taking, helping to increase the efficiency of everyday processes. Boring use cases, for want of a better description. The more exciting ones will follow once users fully understand what Gen AI can deliver.
The same is true in life sciences. Gen AI is being used as a translator in doctor’s offices, to predict disease outbreaks, and to analyze medical images like X-rays and MRIs. It’s even being used to review investigator invoices and automate financial reconciliations—a monthly headache every contract research organization, sponsor, and clinical trials operator will be more than happy to resign to machines.
As the technology develops, however, it will offer more opportunities to expedite drug discovery processes, streamline clinical trials, and speed up regulatory approvals. The McKinsey Global Institute (MGI) estimates that Gen AI could generate between $60 billion and $110 billion annually in economic value for pharmaceutical and medtech organizations––though MGI doesn’t put a date on that figure.
That hesitancy is warranted. Pharma is a conservative industry and will rightly approach something as powerful as AI cautiously. Yes, we’ve leveraged models like QSAR and MoLeR in research to improve drug design and development, but Gen AI applications are not yet commonplace. Integrating AI tools into complex workflows will require time and understanding.
Challenges and Opportunities in Life Sciences
As an industry, we lack the fundamental building blocks to support AI development. It requires large, structured datasets to train and run models, and pharma doesn’t have them. For most, embracing this technology would first mean years of sorting through unstructured and semi-structured data, then cleansing and labeling it.
We’re also insulated from issues like labor shortages that have driven other industries to seek AI efficiency gains. AI’s margin gains simply aren’t as big an incentive for life sciences businesses to embark on multi-year data engineering projects.
Yet, the opportunity extends beyond simply increasing profitability. Within the clinical trials environment specifically, AI will reduce costs and speed up the trial process, facilitating the development of more drugs and increasing access to them. MGI puts the value of this contribution to clinical development at $15 billion to $25 billion.
The problem is that the large networks of small businesses that make up the clinical trials sector cannot support the development of Gen AI for industry-specific use cases. Clinical trials have a different set of problems than other life sciences businesses—Gen AI solutions need to be able to support an operation focused on engaging, recruiting, raising awareness, and randomizing patients into studies.
A Blueprint for Adoption in Clinical Trials
That’s where large integrated site organizations come in. These businesses have the scale to deliver a blueprint for Gen AI adoption in a clinical trial environment. One of the largest such businesses today has access to almost 100 locations globally, and more than 220 investigators. This means it, and others like it, can develop and test Gen AI solutions for clinical trial businesses, putting the technology into everyday use at sites worldwide.
The largest businesses of this sort may also have access to more than 1 million patient records, which provides a sizable training dataset for developing large language models (LLMs). LLMs are now routinely used to extract medical conditions, allergies, and medications from patient histories, converting unstructured, non-standard data into labeled, structured datasets to automatically match patient conditions to study criteria.
But it’s still a relatively immature—boring—use case. The real innovation comes in more complex patient recruitment applications. Gen AI can deliver information in the format that best suits individual users, serving different languages, or providing written information as spoken word. It can simplify the pre-screening survey process, combining different wording for the same questions into one so participants only need to answer the question once for multiple studies.
It’s a leap that many disparate, small clinics working independently would struggle to achieve for years. Yet, integrated site operators have the resources to do it now—and then roll it out to the rest of the industry.
The Impact of Using Gen AI
While these achievements might seem modest in the world of Gen AI accomplishments—especially against grandiose projections of its potential value in the future—they are tangible. And, when it comes to new tech, slow and steady progress is more meaningful than outright revolution. People trust what they’re familiar with, and trust will be critical to the successful adoption of Gen AI in life sciences.
Contributed by Paul Evans, President and Chief Executive of Velocity Clinical Research and former Chair of the ACRP Board of Trustees.
