Using Lexacom's AI Product in Surgery
Did my first day using Lexacom’s ambient AI product live in a real surgery yesterday - does it need a name? (what about Amber? - lol)
Previously I’d largely demo’d it. This was raw and in the wild.
First thoughts - the basic speech recognition without the ambient AI has improved massively - I actually didn’t find it making a mistake. It also cut out come of my mumbling and corrected my grammar!
The ambient AI is almost unreal to behold and I’ve been playing with things like chatGPt for a while now. It listened to numerous 10-15mins conversations and summaries them really well.
Was it perfect - no… but then as I’m being paid to help improve the product I would say that else I’m out of a job but wow… I normally write perhaps too short a consultation record and this not only better recorded the patients presenting complaint and history it did a really good job of working out what my differential diagnosis was and what my action plan was.
My initial comment from doing it live - you perhaps do need to know its listening and speak out your exam findings and work out a way to make this sound natural. For example listening to a chest - it cant hear what you can.. maybe one day! but either say to the patient im just telling the computer my findings or tell the patient your findings.
I can also see that for those that write massive consultations it could really speed up their work.
I also found it useful to summarise to the patient at the end - which I kind of do anyway - something like “so I think you have X, and our plan is to do….” but this isn’t essential. I had one patient with about 5 issues and it got all of them and what I was doing with each.