Using AI to Fight Insurance Denials

1

I saw a segment of the evening news about using AI to fight insurance denials that I found interesting.

Here’s what ChatGPT had to say about with AI programs are best for this:

" :one: Types of AI Tools Available

A. Appeals Drafting & Analysis AI

  • Helps draft appeal letters and analyze denial reasons.
  • Some also scan EOBs (explanation of benefits) and denial codes to match the correct arguments.

B. Claim Tracking AI

  • Monitors claim status and flags denials automatically for rapid appeal.
  • Can detect patterns of improper denials.

C. Legal AI Tools

  • Used by attorneys to cite case law and insurance regulations automatically for stronger appeals.

:two: Best AI Programs Currently Available

For Health Insurance Denials:

:small_blue_diamond: Resolve (getresolve.com)

  • Uses AI + human advocates to appeal medical insurance denials.
  • Handles out-of-network bills and denial appeals.

:small_blue_diamond: Health AI Services in Navina, AKASA, and Olive AI

  • These are typically used by providers to fight payer denials automatically.
  • If you are a medical office or hospital, these reduce denial rates and automate appeals.

:small_blue_diamond: Denial Management Modules in RPA Tools

  • Tools like UiPath and Automation Anywhere can be trained with custom AI to appeal denials in bulk.

For Auto & Homeowners Insurance Denials:

:small_blue_diamond: DoNotPay (donotpay.com)

  • Offers automated appeal letters for denied claims, small claims filing, and regulatory complaint letters.

:small_blue_diamond: Justly

  • AI platform designed to handle insurance disputes and generate demand letters, though more oriented to auto claims and property damage disputes.

Legal AI:

If you are working with an attorney or handling complex, high-dollar denials:

  • Harvey AI (for legal drafting and analysis).
  • Casetext CoCounsel (now under Thomson Reuters, can help prepare legal briefs for denial litigation).
  • Spellbook (for contract analysis if denial involves policy interpretation).

:three: Considerations Before Using These Tools

:white_check_mark: Confirm your policy and denial reason precisely. AI is only effective if your inputs are accurate.
:white_check_mark: Know your appeal deadlines (often 30-180 days).
:white_check_mark: AI is a tool, not a guarantee; sometimes you may still need a human advocate or attorney, especially for repeated or complex denials.

:triangular_flag: Key Tip:

Start with DoNotPay if you are an individual consumer needing a simple letter,
or Resolve if it is a healthcare denial you want handled automatically.

For medical providers, consider Navina, AKASA, or Olive AI.
For law offices, Harvey AI or Casetext CoCounsel will streamline complex denial litigation."

Insurance companies have been accused of using AI to generate denials.

Well, quid pro quo’‘’‘’

1 Like
2

It has been shown there are ways to determine if there is an A.I. being used. Thus A.I. may eventually run out of valid excuses and responses. It is like putting computers against each other playing chess. After a while they determine there are few new moves.

Currently some A.I.s will create text using an em-dash “–” because it is often used in many items used to train A.I.. Personally I have never used an em-dash and I’ll bet most people don’t in normal writing.

I have been watching many Youtube videos and can tell in seconds that they are A.I. generated. The A.I. voice will mispronounce common words and no one is reviewing the video to see the obvious problem. One is when it pronounces a persons name one way, then mispronounces it later on. So right now, if a human can tell it’s A.I. certainly a computer can.

1 Like
3

I agree with Lava but think that it should be pointed out that AI and automation in the work force using AI is in it’s infancy and it will improve as it develops over time.

There are many comparisons to point out but one that was just demonstrated for me is the difference between cardiograms over the last 50 years.

When my father had a cardiogram he went to the doctor’s office and the device was several feet wide and spread out over a large table.

It produced a cardiogram that was valid for that moment.

I recently a a cardiogram by a device that was about the zize of a quarter and was taped on my chest for a week.

It monitored 24 hours a day and was then returned to the manufacturer that down loaded the data and sent it to my doctor.

As for the use of Robots the following news story is illustrative:

Amazon will soon employ more robots than humans as 1 million machines toil across facilities: report

I know that I will not see fully automatic drivers cars in my life time but I am confident that they and driverless trucks will be available in the near future.

I found the following a bid scary:

GPT-4 Passes the Bar Exam: What That Means for Artificial Intelligence Tools in the Legal Profession

1 Like
4

The problem with AI in its current stage of development lies in the computers it runs on.

That problem will probably be overcome when quantum computers become a reality.

Until then, when using AI in it’s current form for complex problems, you stand a chance of looking and sounding like Bugs Bunny.

1 Like
5

Then again there was the case of Mike Lindell’s lawyers using A.I. for briefs and A.I. got all confused and cited inappropriate cases.

I have found ChatGPT creating diagrams and images that are just completely wrong. It gets pretty silly in its solutions too. Yes, it will get better, but you have to consider that right now it is learning based upon information it found on the Internet. A lot of that data is wrong. If you correct it by pointing it to a website containing information it does not already have, it will use that information as correct until later proven wrong. I envision people might fake it out as they have done in Wikipedia.

I uploaded a photo to ChatGPT and it ixed it. On the send fix, it started changing faces.

I went to a website and uploaded a short audio clip of my voice. I had it create an audio of a text entry I wrote. I thought it sounded like me but my Sister could tell it wasn’t after listening for just a couple seconds.

I am tempted to download a version to a local computer and have it real all my blog entries (hundreds of them) and have it analyze me and my writing. I would love t othen give it some ideas for a short story and have it write one for me.

1 Like