Obstacles to AI Deployment in New Product Development
Image: Depositphotos
While excitement about generative AI is high, some questions persist as to how much value is being delivered. AI has been used by leading firms such as Amazon and Netflix to improve shopping recommendations, but examples of significant applications to improve overall business performance are not abundant. One area where AI has considerable potential is new product development (NPD). The NPD process has not changed much in most organizations for decades with fewer than 30% of new product projects becoming commercial successes. Yet only 13% of firms are using AI in NPD.(1)
While there are several barriers to effective AI implementation,(2) the top few obstacles in deploying AI to transform NPD are a bias towards incremental improvement, lack of cross unit collaboration, and challenges with data.
The widespread use of proof-of-concept studies and pilots in AI implementation is one indication of a bias towards incremental improvement. Most pilots and proof-of-concept studies – perhaps up to 90% – are not likely to move into production. While this approach may be acceptable in developing chatbots or yet another shopping recommendation algorithm, it is certainly not enough when it comes to major value creating processes such as new product development. Incremental improvement, while necessary, is not sufficient for success in NPD.
The tendency to deploy AI to resolve small problems inside departmental boundaries is understandable given that it’s new technology. Many companies are still just experimenting with AI at the departmental level. In an Amazon Web Services (AWS) survey only 6% of companies had any production application of generative AI, and in a Wavestone survey only 5% had any production deployment at scale.(3) Management mindset is at the root of such caution. Risk aversion is the principal culprit. Focusing on small applications of AI is widespread. Yet, what’s needed is a big picture view of the potential of AI in transforming end-to-end processes such as NPD.
Challenges with data continue to persist, such as curating unstructured content, improving data quality, and integrating diverse sources. In the Amazon Web Services (AWS) survey, 93% of respondents agreed that data strategy is critical to getting value from generative AI, but 57% had made no changes to their data thus far.(4) Despite the excitement around AI, current AI technology still has some considerable problems. It cannot reliably deal with facts, perform complex reasoning, or explain its conclusions.”(5) Data plays a pivotal role in AI deployment. These challenges represent significant obstacles in applying AI to NPD.
The potential of AI to compress cycle time in new drug development is particularly noteworthy. The statistics are remarkably dismal for NPD in new drug development where the overall probability of success for all drugs and vaccines is 13.8%.
The potential of AI in the oncology clinical trial process is particularly interesting. It typically takes over 10 years and costs over a billion dollars to bring a new cancer drug to market. Cancer drug development has had the worst performance at a 3.4% success rate across all therapeutic areas.(6) Some observers have referred to Eroom’s law which states that the inflation-adjusted cost of developing a new drug roughly doubles every nine years and noted that the cost of developing a new drug has increased exponentially in the last several decades despite improvements in technology.(7) The obstacles of a bias towards incremental improvement, challenges with data and a lack of collaboration across departments are especially worrying in the oncology clinical trial process. Only a handful of established companies are deploying AI and data-driven approaches systematically in their clinical development – and fewer still are deploying AI systemically.(8) The high level of risk averseness in the oncology clinical trial process is due to both the extremely regulated nature of the pharmaceutical sector as well as the traditional mindset of leaders in big pharma.
How to overcome these challenges? To take best advantage of AI, C-suite executives in large pharmaceutical companies need to make a genuine commitment to deploying AI on a large scale in clinical trials. That means looking at the potential of AI in the context of the end-to-end clinical trial process – not just in one part of it. While current practices in patient selection and recruiting mechanisms do require attention, so does the inability to monitor patients effectively during oncology clinical trials.
Likewise, a concerted effort is needed to deal with existing data quality and curation challenges including, but not limited to, redundant, outdated, and conflicting information. This may include the purchase of certain commercial data sets, such as electronic health records and claims, and other types of data sets—registries, biobanks, and clinical trials from other pharma companies. Focused attention is needed to improve interoperability and establish standardized electronic health record (EHR) formats.
Given the rapid evolution of AI, biopharmaceutical companies might also consider building their own AI capabilities by attracting top data scientists and data engineers.
Due to the highly regulated nature of the clinical trial process, perhaps the most critical undertaking is for C-suite executives in large pharmaceutical companies to work tirelessly with internal departments and external regulators for an unprecedented level of collaboration.
AI can not just be used to inform clinical trial eligibility criteria, enhance the diversity of participants, and reduce sample size requirements. It should also be implemented to create an external control arm to make trials more patient-centric, shorten enrollment timelines, and increase statistical power. AI may also be deployed in combination with smart devices, such as wearable sensor devices, to develop efficient, mobile, real-time, constant, and personalized patient surveillance systems that can monitor patients more effectively during the trial period.(9)
Several leading pharmaceutical companies have recognized the potential of AI to improve NPD performance.(10) The success of future oncology clinical trials requires a fundamental transformation in how trials are designed, conducted, monitored, adapted, reported and regulated to generate the best evidence.(11) Status quo practices are unsustainable. Silos need to be broken. AI has the potential to enable a new model for preventive, personalized, pragmatic and patient-participatory clinical trials. What’s needed now is to aim higher and chose to cure cancer!
Applying AI to the oncology clinical trial process is particularly important as millions of lives are at stake. The key principles also apply generally to NPD - focusing on the entire process - not just one part, breaking down silos, encouraging collaboration, and overcoming data challenges.
Andrew Spanyi is President of Spanyi International. He is a member of the Board of Advisors at the Association of Business Process Professionals and has been an instructor at the BPM Institute. He is also a member of the Cognitive World Think Tank on enterprise AI.
Cooper, Robert G. "The Artificial Intelligence Revolution in New-Product Development." IEEE Engineering Management Review, Vol. 52, No. 1, February 2024
Marr, Bernard. "11 Barriers to Effective AI Adoption and How to Overcome Them." Forbes, May 10, 2024
Davenport, Thomas H., and Randy Bean. "Five Key Trends in Data Science." MIT Sloan Management Review, January 9, 2024.
Ibid
Berezow, Alex. "Clinical Trial Success Rates by Phase and Therapeutic Area." American Council on Science and Health, June 2020
Zhavoronkov, Alex. "When Will AI Beat The Eroom’s Law In The Pharmaceutical Industry?" Forbes, August 2022.
Buntz, Brian. "How 11 Big Pharma Companies Are Using AI." Drug Discovery Trends, June 15, 2023.
Subbiah, Vivek. "The Next Generation of Evidence-Based Medicine." Nature Medicine, January 2023