Opinion: Can AI Clinical Trial Technology Deliver on Its Promise?

Over the past 30 years, the clinical trial ecosystem has undergone remarkable technological transformation, such as the transition away from manual data entry to electronic data capture. Yet many organizations have struggled to translate technological excitement into meaningful operational efficiency. The time it takes to get final results is the same or longer as it’s always been, and the breakthrough rate for new drugs remains about the same as it was 30 years ago.

As a senior director at eClinical Solutions, which provides clients with an AI-enabled clinical trial data platform, I don’t need to be convinced of the potential of machine learning to facilitate drug development. But I fear that if not implemented correctly, AI will, like previous innovations, fail to move the needle on the time and expense needed for drug development.

The volume of data collected in Phase III pivotal trials surged by 283% between 2010 and 2020, according to the Tufts Center for the Study of Drug Development. This dramatic increase reflects growing clinical trial complexity, with more intricate protocol designs and a greater burden of data collection. Tufts also found that as protocols become more complicated, clinical trial performance tends to suffer, as evidenced by longer timelines, higher costs and declines in data quality. Bringing a new drug to market now costs an estimated $300 million to nearly $4.5 billion, increasing at an annual rate of 7.4%.

Thus, despite technological progress, developing a drug is no quicker or cheaper than it used to be. Unless we understand why, we risk seeing this trend continue despite the current enthusiasm for adopting AI.

Ten to 15 years ago, the industry embraced new technologies like blockchain, virtual reality and robotic process automation that backers purported would accelerate drug development, reduce costs and address the biggest life sciences challenges. But today’s reality is clear: Despite the big promises and investment, companies were not always seeing return on investment (ROI) at scale.

One reason is that implementing new technology alone isn’t enough to drive meaningful change. It certainly plays a key part, but as recent data reveal, organizations need the right processes and people in place if they want to succeed. Indeed, a common pitfall is to take a tech-first approach, when change must instead start with the people and teams that will adopt these technologies. Moreover, rolling out new technology must come in the context of the long-term goals of the company. Only with such a holistic approach can companies adapt their existing processes to fully realize the technology’s potential.

It is imperative we do not repeat the mistakes of the past when applying AI to our daily operations. We must remain focused on business impact—and on the processes, capabilities and attitude needed to achieve it—rather than inflated expectations.

Too often, the industry considers technologies to be the goal in and of themselves, when instead tech should support the end goals of improving care and outcomes for patients by bringing new therapies to market. When adopting AI applications, companies must define targeted, specific outcomes that make sense for the organization. One size does not fit all.

Lofty goals of increasing productivity by 75% at 30% of the cost must be supported by the reality of current benchmarks, data-driven opportunity projections, alignment across functional boundaries, desire (or not) to change operating models and appetite to truly transform.

In order to harness technology as a means to our ultimate goal of supporting patients, we first need to get better at defining targets for that particular initiative. Then, we must scrupulously hold ourselves accountable to those objectives over the next few years. In my experience, these steps are routinely missed as conversations focus largely on the technology rather than what it is actually going to do.

It is key for companies to address these challenges before implementation, starting with identifying the specific problems for technology to support. This means examining the current operating model to identify areas where speed and productivity can be massively improved.

A large part of my role in customer success and strategy is to help life science leaders realize real, measurable outcomes. I’ve learned that if you are not asking the right questions upfront, it will likely affect the outcome. Without understanding your KPIs today and over time, you risk deepening the disconnect between innovation and genuine industry value.

Organizations and teams must clearly define KPIs to measure and recognize corresponding value. AI has the potential to be transformative, but as such it’s even more imperative to start with defined use cases tied to measurable KPIs. As one example, productivity KPIs will highlight whether AI is making processes better or adding friction, so that barriers to adoption can be identified early. KPIs tied to business objectives, such as cycle times in a specific area or quality metrics, prevent the overinflated results of AI from falling into a trap of nebulous definition of success leading to nebulous expectations. Metrics are essential to help assess whether an implementation’s KPIs have been realized, and more importantly, if we are actually moving the needle for patients.

AI holds great promise in overcoming today’s clinical trial obstacles. But if we repeat past mistakes, that won’t happen. Lasting changes are not achieved simply by selecting the latest hyped technology, however powerful it may be. I challenge teams to ask themselves: What must we change? As AI continues to advance and evolve, drug development cycle times should not remain stagnant. If this time we reinvent the approach, we may finally unlock transformational results.