What your Medicare plan isn't telling you

Today's Tip: Prep for Medicare Open Enrollment

Medicare Open Enrollment opens in 8 days. And most people are about to make the same expensive mistake they made last year… because they never looked at the one piece of mail that tells them it's coming.

Let me explain…

Every September, your Medicare Advantage plan or your Part D drug plan is required by law to mail you a document called the Annual Notice of Change… the ANOC. It's usually a thick envelope. Maybe 20 or 30 pages. Dense. Boring. The kind of thing most people glance at, decide to read later, and never open again.

Inside that envelope is the single most important piece of information you'll get all year about your health coverage. It lists everything that's changing about your plan for 2027… new premium, new deductible, which of your drugs got moved to a worse tier, which of your doctors dropped out of the network, which pharmacies are no longer "preferred."

Take my neighbor Margaret. She's 68, retired bookkeeper, lives two doors down. Last year she tossed her ANOC in the "deal with it later" pile. January rolled around and she went to refill her blood pressure medication. The co-pay had jumped from $12 to $87. Her plan had quietly moved it from Tier 2 to Tier 4… and it was all right there in the ANOC she never read.

Here's the thing. Reading the ANOC yourself is miserable work. The language is written by insurance lawyers for insurance lawyers. But AI can read it for you in about two minutes and tell you — in plain English — exactly what's changing and whether it matters for your situation.

Here's the prompt that does it…

How do you actually upload the ANOC? Easiest way: take a clear photo of each page with your phone and upload the photos directly in Claude or Gemini. Both can read scanned pages. If your ANOC came as a PDF by email, even better — drag it right into the chat window.

What you get back is a one-page cheat sheet you can read in under a minute. Premium up or down. Deductible up or down. Your medications flagged if they moved. Your doctors flagged if they left the network. That's it. The 30-page document becomes a 10-bullet summary tailored to your life.

Now, I know what you're thinking… "I don't want some chatbot making my Medicare decisions for me." Good. You shouldn't. And this prompt doesn't ask it to. It asks the AI to do one job: read a long document and tell you what it says. That's the same job a trusted friend who happens to be good at reading insurance paperwork would do for you… except AI doesn't get tired on page 14.

Once you know what's changing, THEN you make the decision. And for that part — picking a plan — you go to the official Medicare Plan Finder at Medicare.gov and run your actual drugs and doctors through their tool. The AI reads the mail. Medicare.gov quotes the prices. Together they do in an evening what used to take a week.

So here's your assignment this week. Go find the Annual Notice of Change your plan mailed you last month. It might still be on the kitchen counter. It might be in the pile by the door. Pull it out. Run it through the prompt above. Open enrollment starts October 15… you have a few days to be ready instead of surprised.

Two minutes of AI could save you a thousand dollars. Not a bad trade.

★ Quick Wins

Skill Builder: Catching AI Hallucinations

Here's a word every AI user should know: hallucination. It's what happens when an AI confidently tells you something that is completely, 100% made up. Not a typo. Not a small mistake. An invention. A doctor's name that doesn't exist. A phone number for a business that was never real. A court case citation a lawyer in New York got sanctioned for using last year because the cases didn't exist at all.

The frustrating part is that AI hallucinations don't look wrong. They look polished. Grammatically perfect. Full of specific-sounding details. That's exactly what makes them dangerous for the 55-and-up crowd, because this generation was raised to trust confident-sounding writing.

Three categories trigger hallucinations more than any others: specific numbers (prices, dates, statistics), proper names (people, companies, book titles), and anything that happened recently. If you're asking about any of those three, your skepticism should go up.

The best defense is a prompt that gives the AI permission to say "I don't know." Most people don't realize this, but AI systems try to be helpful to a fault — they'd rather guess than admit uncertainty. You have to explicitly tell them it's okay not to know. Here's a short template that works beautifully:

Save that prompt somewhere you can grab it fast — a sticky note on your monitor, a note in your phone. Paste it in anytime you're asking an AI about something that actually matters: medical questions, legal questions, anything with a price tag, anything you'd be embarrassed to be wrong about in front of your kids.

The habit you're building is small but powerful: treat AI like a brilliant intern. Fast, hardworking, often right… but you still read what the intern writes before you sign your name to it. That's not pessimism. That's how smart people use AI.