Generative AI Use Cases: Part 1

Justin Yu, Chad Patel

Generative artificial intelligence (AI), a type of AI that creates new content such as text and images, has applications throughout the drug commercialization process. Currently, however, limited information and examples are available regarding optimal use cases in pharma. In this article, we will discuss six use cases for generative AI in relation to Value & Access. Most of these use cases are also applicable to other related functions (e.g., medical affairs, medical information), and we encourage you to experiment with AI tools available in your organization to identify areas with the highest return on investment. As with many generative AI use cases, optimizing your AI prompts typically require multiple rounds of iteration. The better you understand your input data (e.g., format, data structure) as well as strengths and limitations of different AI systems (e.g., ChatGPT, Claude, Gemini), the more effective you will be at generating AI-based solutions.

Use case 1:

Generating FAQ documents

Generative AI excels at summarizing text, but sometimes requires additional instructions to fit the needs of the user (e.g., inclusion of key points that may be relevant to only certain customer segments, use of plain language instead of medical jargon, inclusion of citations). Different AI systems (e.g., ChatGPT, Claude) may also produce different results (e.g., the two AI systems identify different questions for the FAQ document), so we encourage use of more than one AI system when refining any AI-based solution.

Interested in optimizing the development of a FAQ document or objection handler? AESARA’s expertise in scientific communications will ensure that any solution created with the assistance of generative AI meets your needs. Please contact AESARA to learn more.

Use case 2:

Recreating tables from images

Generative AI’s ability to recreate scientific tables from images has the potential to save researchers many hours of tedious work over the course of a year.

Exercise: In the following examples, ChatGPT does not produce tables in a usable format (e.g., Word table, Excel table). How would you modify the prompts to generate a more desirable output (hint: You could also try Gemini, which has the ability to output results directly to a Google Doc)?

Interested in learning more about how to best deploy generative AI to synthesize key images into meaningful tables across the literature, dossiers, and other internal documents? Please contact AESARA to learn more.

Use case 3:

Creating a chatbot for answering questions related to specific content

This example is based on a recent webinar (Review of FDA Warning & Untitled Letters: Real-Life Examples & Insights) by AESARA and OpusRegulatory on March 29, 2024. One limitation to this no-code approach is that custom GPTs are currently available only to ChatGPT Plus subscribers. Other options include programming a chatbot powered by an application programming interface (API) (e.g., OpenAI, Anthropic) or a locally-run large language model (LLM).

Interested in creating an AI-powered chatbot to help answer questions from internal and external stakeholders? Please contact AESARA to learn more.

Use case 4:

Summarizing data from articles for use in dossiers

The example above was based on a tabular format with column headers “Citation,” “Design, Sample Size, Treatments,” “Inclusion and Exclusion Criteria,” and “Endpoints, Results.” Once you provide details on the format of choice for displaying data, generative AI can do the rest. Note how ChatGPT and Claude align on most generated content, but differ in some areas (e.g., Claude directly stated that study participants were randomized 1:1:1).

Exercise: Can you think of how you can adapt this approach to summarize articles for Section 3.1B (Study Summaries) of the AMCP dossier? Can you generate better results the first time around by providing specific instructions versus providing “gold standard” examples?

Interested in shortening the development time for pre- and post-approval dossiers? AESARA can help by providing expert guidance on optimal formats and language for value communication. Please contact AESARA to learn more.

Use case 5:

Generation of plain language summaries of articles or other text

Depending on the audience (e.g., average lay person, medical professional), the reading level of the plain language summary can be adjusted accordingly. Your choice of AI system (e.g., ChatGPT, Claude) can also result in different variations of a plain language summary of text. In the following examples, Claude’s initial output using the same prompt as for ChatGPT resulted in a plain language summary that was more conversational in nature and at a lower reading level compared to ChatGPT’s initial output.

Exercise: As an exercise in developing plain language summaries, try “training” an AI system using published peer-reviewed plain language summaries from a target journal and see how results differ from the given examples. Do the results improve if you keep adding more training examples? 

Interested in creating a plain language summary of an article? AESARA can help you optimize the output from generative AI by identifying “gold standard” examples (e.g., from top tier journals) and relevant guidelines (e.g., best practices for plain language communication). Please contact AESARA to learn more.

Use case 6:

Creating custom images to support communications

Generative tools like Midjourney can be used to enhance communication of data, capture the attention of target audiences, and support stories that show extremely specific scenarios.

Interested in creating custom images to support and create memorable internal and external communications? Please contact AESARA to learn more.

Generative AI is already making an impact in the life sciences industry, especially with regard to the identification of new compounds for the potential treatment of various diseases. With regard to the use of generative AI beyond clinical development, however, a recent ZoomRx survey found that 65% of the top 20 pharmaceutical companies currently ban or restrict the use of ChatGPT. At AESARA, our technical experts understand the importance of keeping your data safe and applying generative AI in a secure environment. See our ATLAS application for additional details. Our experts in Value & Access also encourage you to try the above use cases and share your thoughts on additional use cases that AESARA can help you think through and/or implement.

Contact us

Fill out the form below, and we will be in touch shortly.