AI in Healthcare 🤖

We’ve dabbled a bit, but haven’t made any sweeping changes. I do love the Sprout AI feature for writing social copy, it’s a great way for me and my team to brainstorm and get “unstuck” when inspiration is low.

Hey Anna! Totally heard. For the healthcare side of the house, do you think that’s because leadership is cautious to adopt AI for potential security/data issues (even though we’re seeing that AI is quite safe at this point)? Or something else? @Anna Laura McGranahan 

Came here to say the same! We are still working on security and data concerns for AI as well. 

My team has done some testing of Generative AI for responding to customer reviews. The challenge is the AI tries to be TOO specific, sharing more info in the response than HIPAA allows. If someone writes about what a great experience they had delivering their baby at our hospital, the AI infuses the response with things such as “Congratulations on your new baby!” So until there’s an LLM specific to healthcare that plays by HIPAA and other privacy rules, or we create and train our own internal LLM, it may take more time for agents to edit the AI responses before posting, making it less of a time savings. 

Thank you, Julie! Has your team’s “dabbling” mainly been for content creation/ideation, or something else?

From a security perspective -- is the main concern that an AI-based system (such as ChatGPT) would be connected to one of your internal systems where patient information is held? @Julie Cochran 

@njperry Did you just come up with the next big AI company (HIPAA compliant AI)?! Count me in. 👀 Haha, I’m sure someone’s already on it, so will be interesting once industry-specific AI models launch. In the meantime, has your team thought about using canned responses or saved replies as a starting point to help guide the AI as to what it can/can’t say in a message?

Mine is a little different as I’m in the nonprofit research world, but we are dabbling. That said, I am only using it for tweak social copy that I have already written for different platforms. It has also been great when I get stuck writing my monthly Twitter/X chat to form it into more of a narrative where each question builds on the other and there is less back and forth when having it medically validated.

Right now, just for idea generation and proofreading. Things like Chat GPT are blocked on work devices… so is social media for many employees, too. Lots of hurdles to jump through...