The famous quote attributed to Voltaire, which claims that “perfection is the enemy of the good,” is rarely applied in reference to artificial intelligence. However, as increased concerns and criticisms surface regarding this new approach to information analysis, the life sciences industry must remind itself that, as organizations heavily reliant on the aggregation of information, we cannot disregard emerging technologies simply on account of imperfection.
Throughout history, disruptive technologies have consistently been met with resistance, culminating in a gradual transition over time. For example, although the first light bulbs were invented in the 1870s, by the early 1900s only 5 percent of manufacturing processes utilized electric power and even continued to rely on steam power until the late 1910s. The introduction of novel technology is likely to be met with hesitancy, but with proper guidance, these new developments maintain the potential to mitigate industry-wide challenges.
Data management challenges in the life sciences industry
Across the life sciences industry, stakeholders admit that the growing influx of unstructured data presents a major challenge. Currently, most pharmacovigilance data processing activities are performed manually with the help of large databases. These activities include case management, signal management and aggregate reporting. While some of these activities have been automated with the advancement of technology, these classic automations have been widely exhausted, and organizations still struggle to manage the increasing volume of adverse events. Besides the influx of data in the pharmacovigilance landscape, organizations are also struggling to keep up with increasing regulatory requirements, which forces them to constantly revise and confirm that they are adhering to the latest guidelines. Combining these factors with the ongoing shortage of qualified labor continues to impact the field and create additional challenges regarding workload.
To address these challenges, many organizations are turning to automation and increasingly view generative artificial intelligence (GenAI) as a method to improve the efficiency of manual tasks and glean valuable information from an overwhelming amount of data. Though the use of GenAI carries both advantages and disadvantages, when managed and trained correctly with proper oversight, organizations can mitigate the potential risks and benefit from the time-saving aspects of this technology.
Life sciences’ hesitancy to implement GenAI
Understandably, pharmaceutical companies are wary of the risks associated with GenAI, such as potential bias, lack of reliability and overall mistrust in the validity of data and outputs. These concerns are legitimate, and issues like data privacy continue to circulate in conversations regarding this technology. As a result, organizations are approaching the implementation of GenAI processing cautiously to not inadvertently expose patient data. Similarly, many companies maintain concerns about data quality and are aware that the effectiveness of GenAI is entirely dependent on the quality of the data fed into the system. Finally, concerns remain that undue disruption of established processes with the implementation of this new technology would reduce productivity and that the rising costs of this technology may not justify the investment.
Though the concerns regarding GenAI are understandable, the industry cannot deny the benefits of this technology in other areas. Experts suggest that approximately 50% of life sciences work hours will either be automated or augmented in the future with the help of this technology.
Value of GenAI use in pharmacovigilance
Implementing GenAI into processes provides many benefits, such as code creation, data summarization and the acceleration of current artificial intelligence (AI) applications. Specifically, in pharmacovigilance, GenAI can compile data, convert inbound unstructured data into structured data and create a first draft of requisite document narratives.
It can also provide great value in improving clinical efficiency and outcomes by providing early signal detection. Essentially, any situation that requires the processing and analysis of large amounts of data in a timely manner can benefit from GenAI. Allowing this technology to take on the time-consuming, repetitive work traditionally performed by humans allows us to focus more time on the more valuable aspects of managing the drug lifecycle.
Ensuring proper use of GenAI
Organizations must manage the challenges and risks associated with the use of GenAI. The key to successful implementation of this technology lays in proper human oversight and continuous validation and retraining of algorithms, otherwise known as “human in the loop.” In order to benefit from the value of GenAI in pharmacovigilance workflows, organizations must keep in mind the following considerations:
- Specific use case: Ensure that the implementation of GenAI solves a specific, practical problem. Identifying a specific use case creates focus and provides a legitimate business case for the investment of both time and money.
- Data quality and standardization: To fully leverage GenAI, organizations must collect and standardize data in a way that can be easily understood by machine algorithms.
- Integrate data scientists: Involve experts in conversations about what problems this technology is trying to solve.
- Consider regulatory compliance: Regulators are doing their best to keep up with GenAI and face the same challenges that enterprises do in balancing compliance with embracing change.
- Ensure continuous falidation: Practice validation to confirm that outputs align with their intended outcomes.
- Train humans on prompt engineering: Educate employees to better understand how to prompt GenAI and ask the correct questions that will provide desired answers.
- Change management: Work across leadership groups to generate momentum and provide an understanding of the value of GenAI.
- Human in the loop: Maintain control of machine algorithms by requiring continuous human oversight to mitigate risk.
Despite concerns regarding GenAI and its reliability and safety, many expect it to make a significant impact on the life sciences industry, with 90% of biopharma and medtech respondents expecting GenAI to have an impact on their organizations within the year. Still, lingering hesitancy remains, as 25% of medtech executives and 18% of biopharma executives claim to prefer to wait for more evidence to emerge before implementing the technology.
Considering the lessons of history, we can expect that GenAI will eventually be embraced throughout the life sciences industry. Understanding that GenAI is not perfect and will require continuous training, oversight and revision is part of the process of implementing new technology. The same is obviously true with humans. Though imperfect, GenAI can still provide valuable insights and substantial improvements to the compilation of data and the generation of crucial documents. Disregarding novel developments on account of imperfection will postpone the advancement of scientific discovery. The key to successful implementation of GenAI is ensuring appropriate guardrails and guidance, involving humans in every step of the journey.
Photo: Quardia, Getty Images
