What the agent did
After his uninsured brother in law died following a heart attack, the hospital sent a $195,628 bill. Matt Rosenberg asked Claude to put the billing codes in a spreadsheet and look up what Medicare pays for each. Claude found that the hospital had unbundled a cardiac procedure, billed a bypass that never happened and billed ventilation alongside critical care. He had ChatGPT check Claude's analysis, spot checked the rules himself, then sent a six page letter offering the Medicare rate. Claude also helped draft the settlement agreement.
Result, as claimed
Claude calculated Medicare would have paid about $28,675 instead of $195,628. The hospital countered at $36,356 and they settled at roughly $32,500. "The collaboration with Claude and ChatGPT hadn't just saved $163,000."
This is the result as claimed in the public source linked below. It has not been checked and is not a typical outcome.
Proof
Try it yourself
Ask for the coded bill, then have an AI compare each line with what Medicare or an insurer would pay.
Prompt
Here is an itemised hospital bill with billing codes (UB-04 or CPT): [paste or attach]. Location: [city and state]. Date of service: [date]. Make a table with each code, what it is, the billed amount and what Medicare would pay. Flag any line that looks unbundled, duplicated, inpatient only on an outpatient visit, or not supported by the records. Cite the Medicare rule for each flag. Then draft a short, factual letter offering to pay the Medicare total. Do not send anything.
Tools
- Claude Named in the story. He used it to research Medicare rates, find billing errors and draft the settlement
- ChatGPT Named in the story. He used it to check Claude's analysis for errors
Steps
- Do not pay a large bill straight away. Ask for the itemised bill and the UB-04 or coded claim form.
- Run the prompt, then ask a second AI to check every flag.
- Read the actual Medicare rules for each flag before you rely on it.
- Send the letter, negotiate, and get any settlement in writing.
Watch out: AI can be wrong about codes and rates. Check every figure against official sources, and for large sums talk to a patient advocate or lawyer before you sign.
Starter version by Solopreneur Code
Skill: Bills and savings
Add this skill
Hospital bill checker (US)
Compares each line of an itemised US hospital bill with what Medicare would pay and flags lines that look wrong.
- Works with
- Claude, ChatGPT, Gemini For US bills with billing codes. The story used Claude, then had ChatGPT check the result. A second AI check is part of the skill.
- Tools in the story
- Claude, ChatGPT
- Country
- United States
How to use this: Copy the skill, paste it into Claude, ChatGPT or Gemini, and ask it to save it as a skill. If it cannot, it will still follow the steps.
Show skill
If the Copy button does not work, select this text and copy it.
Job: check a medical bill and dispute errors
Other tools for this job
Tools used
- ClaudeFree, then Pro $20 a month ($17 a month billed yearly), Max $100 and Ultra $200 a month. Claude Code is included in paid plans
- ChatGPTSee official pricing: ChatGPT, opens in a new tab
Proven in other stories
- A patient used ChatGPT to spot billing problems and write insurance appeals, and never paid a surprise $1,200 dental billChatGPT read a dental bill, flagged wrong charges and drafted insurance appeals and a complaint.
Could also do this
- Counterforce Health Not tested in a storyHelps patients write appeals against health insurance denials with AI and an expert team. Free appeal generator. It covers insurance denials, not checking hospital billing codes.