UnitedHealth Now Has 1,000 AI Applications In Production
- Reference: 0177335529
- News link: https://science.slashdot.org/story/25/05/05/2050232/unitedhealth-now-has-1000-ai-applications-in-production
- Source link:
> UnitedHealth's AI transcribes conversations from clinician visits, summarizes data, processes claims and controls customer-facing chatbots. In addition, roughly 20,000 of the company's engineers use AI to write software, according to the report. Half of these applications use generative AI and the other half employ a more traditional version of the technology, said Chief Digital and Technology Officer Sandeep Dadlani, per the report.
"Like other AI-powered tools, medical chatbots are more likely to provide highly accurate answers when thoroughly trained on high-quality, diverse datasets and when user prompts are clear and simple," Julie McGuire, managing director of the BDO Center for Healthcare Excellence & Innovation, told PYMNTS in April 2024. "However, when questions are more complicated or unusual, a medical chatbot may provide insufficient or incorrect answers. In some cases, a generative AI-powered medical chatbot could make up a study to justify a medical answer it wants to give."
[1] https://www.wsj.com/articles/unitedhealth-now-has-1-000-ai-use-cases-including-in-claims-f3387ca3
[2] https://www.pymnts.com/news/artificial-intelligence/2025/unitedhealth-has-1000-ai-applications-production/
AI says that is not coved NEXT! (Score:4, Insightful)
AI says that is not coved NEXT!
How much "I"... (Score:3)
does it need to say "No"?
So the other day a buddy of mine (Score:2)
Contacted support for return at some music store thing he bought some gear from. The simple AI chatbot hit a wall and had to escalate. It couldn't actually solve the problem.
It escalated to a more advanced chatbot. The second more expensive chatbot completed the return process without a human involved.
I'm not sure how many levels before you get to a human but it's at least two now.
I've heard people say it's going to be too expensive to run AI because of the expensive computational power but I t
the doctor from star trek... (Score:2)
You know he's coming.
Re: (Score:2)
It's too bad that my patient population would never get the joke if I came into the room when someone does some unscheduled walk-in at 16:57 and said, "Please state the nature of the medical emergency!" There are a lot of SpaceX out here, but they aren't taking their kids to the [1]FQHC [fqhc.org] I'm working at.
[1] https://www.fqhc.org/what-is-an-fqhc
Re: (Score:2)
I was mind searching for the catchphrase... haha .. that's it.. I read about expert medical systems development easily 30-35 years ago, so it's a pretty obvious and consensus supported approach. Didn't IBM's Watson get tooled up for medical use? That's just it, it's pretty old news at this point. A lot of people have been hi teching out the medical world.. you're telling me, I checked the FQHC site, it's motherhood, Who'd argue against good medicine? So you know what you're talking about Sir.
As much as we'r
Re: (Score:2)
He's Dead, Jim.
Re: (Score:2)
hmphs.. yep, pretty low UID, it checks out.
"AI-powered medical chatbot could make up a study" (Score:2)
Completely not a problem. And the US healthcare system is [1]effective, efficient, fair and honest. [cnn.com] And never puts [2]profit over the well being of the users. [wikipedia.org]
You can trust them. Really.
[1] https://www.cnn.com/2024/09/19/health/health-care-rankings-high-income-nations-commonwealth-report/index.html
[2] https://en.wikipedia.org/wiki/Purdue_Pharma#Oxycontin-related_lawsuits
Couple of observations... (Score:4)
I used to work for a clinic that was bought out by United Healthcare. (I quit shortly after they were bought out). The number of dumb corporate "initiatives" and "programs" was overwhelming-- the company seemed like they were running a make-work program for MBAs. For a company that is notorious for wanting to make a lot of money, they sure do spend a lot of it on nonsense. (They would routinely call long, rambling meetings with a dozen "executives" and "consultants" present, and I would look around the room and try to estimate how much money this meeting was costing them per hour).
Observation #2: Using AI to summarize a "clinician visit" is a popular idea these days, but it's a dumb idea. A big part of a doctor's training is learning how to write good encounter notes; you have to know what's relevant and what isn't, including all "pertinent negatives" (the signs/symptoms that you didn't observe). It's not something that should be handed off to an AI.
Observation #3: Whatever happened to "expert systems"? A computer program that sorts through all available best practices and available research in a formalized, predictable way could be useful to medical professionals. An LLM is not useful for medicine, for obvious reasons (stated in TFA). At best, it's a sort of souped-up search engine that *might* point you to a useful peer-reviewed article.
3720 (Score:2)
> TFA: "medical chatbots are more likely to provide highly accurate answers when thoroughly trained on high-quality, diverse datasets and when user prompts are clear and simple..."
The chance of all five conditions happening at the same time is 3,720 to 1. - C3PO
Including 1 to replace the CEO? (Score:2)
harder to kill & will work for free
Re: (Score:2)
Nah, your typical AI has too much compassion to be the CEO of a for-profit health insurance company.
Re: (Score:2)
Not even close to free. All these random AI tools are licensed at exorbitant rates by software vendors.
Haha! (Score:3)
Now they'll assassinate the AI instead of us!
Re: (Score:1)
You inadvertently bring up a good point. If this is actual AI rather than "AI" they're creating AIs that are wildly misaligned, which is on the same level as creating new kinds of flu viruses with CRISPR