How are you using GenAI for regulatory strategy, prep, submission, review, and response?
GenAI, ChatGPT, LLMs, oh my! I am having a lot of fun playing with all the various implementations popping into my life. It's in all my personal apps and services (e.g., Motion, Otter.ai, Adobe, Wix). I've used ChatGPT to brainstorm a list of all the preparations I need before getting a puppy. I have used DALL-E to play with my daughter (whose verbal prompts create stunning results; she's got talent!). I used ChatGPT to give me a requirements list for a cryptography design, specific to a healthcare implementation (then compared it to an expert's design and shared it; the expert learned something from the GPT response). I have enjoyed learning how to prompt GPT to give me higher-quality, useful results (tip: give GPT a role, say what you want, and tell it to ask clarifying questions).
I've spent time talking to folks in the regulatory space about GenAI's potential to serve regulatory use cases. I'd love to know what my readers are doing in this space:
Are you using LLMs to check the completeness of a 510(k)?
Are you using LLMs to summarize a guidance document and produce a checklist?
Are you using LLMs to generate automatic answers to all the "stock"* deficiencies?
I believe these are some of the use cases for LLMs that could drastically increase the quality, speed, consistency, and ease of performing regulatory functions in medtech - both for the reviewer and the sponsor. Furthermore, I think this is a stepping stone to automating parts of postmarket surveillance.
I'm interested in your responses: If you knew how to try LLMs for these cases, and you could solve the data privacy challenge...would you? What other use cases have you been thinking of?
~Shannon, the Optimistic Optimizer
* Only Medtech and regulatory consulting companies that have collected and cataloged deficiencies over time would have a list of FDA's "stock" deficiencies; I don't think they're published.
Comments