With budget season upon most organizations, the questions about where to invest and how to deliver growth results or realize efficiencies are top of mind. Reflecting on 18 months with generative AI solutions and continued exchanges and discussions about yet more transformative potential of AI for medical, I see three take aways for decision makers.
1. Private generative AI can deliver high quality results for medical affairs
Numerous use cases that we’ve been able to implement with pharma, medtech and biotech companies over the past 18 months demonstrate that reliable results can be delivered. Controlled, private large language model instances and clean data lay the foundation. It is possible to avoid and eliminate hallucinations and have clear visibility and linkage back to the source reference documents. Continued healthy skepticism is warranted, as the participants of a webinar observed last week.
2. Focus on the business pain points
Organizations that focus on the pain points and less on the technology tend to identify more relevant use cases. Whether the issue is rapidly building capabilities without massive investment or hiring, managing peaks and troughs in the business, or capturing efficiencies because more must be done with less – gaining clarity about the business needs sets the more effective companies apart. While 3 criteria are usually sufficient to prioritize the use cases, some organizations prefer to use up to 15. Based on the estimated benefit, the complexity to implement, and the time horizon, the most attractive use cases are then identified. We see companies achieving triple digit ROI and payback periods of 3-6 months.
3. Making technology available is not correlated to results in Medical Affairs
Many organizations are making private instances of publicly available large language models available within their controlled environment. The results are often mixed and of limited value, leading to a slow uptake and missed expectations. As Mayo Clinic learned with their pilot of generative ai that produced mixed results, supporting users by defining a controlled setting and specific tasks is more likely to produce productive results. We see organizations capture more value when clearly defined low to medium complexity tasks such as literature tracking and summaries, lay person summaries, and reference validations are tackled first. This allows people to build expertise and adjust their workflow while building capabilities.
Contact us to discuss how private, generative AI can accelerate your projects in Medical Affairs.