I’m getting (cautiously) AI-curious

Hey there,

Last year, we talked with tech worker Holden Karau, who had built an AI-powered tool to fight denials by insurance companies with the slogan “Make your insurance company cry, too!”

Since then, more tools have cropped up. A few months ago, an episode of The Nocturnists (a podcast made by and for health care workers) looked at a company called Claimable.

So did an NBC News story by the great reporter Gretchen Morgenson, whose clip also mentioned Holden’s tool and a couple of others.

Listeners sent us links to all those stories. Another listener wrote to us about an AI tool he was developing — trained in part, he said, with tips from our reporting — and invited us to test it. (We haven’t yet.)

Then, a few weeks back, some of you sent us links to a social media thread by a user who claims to have used Microsoft’s Claude AI assistant to help knock more than $150,000 off a hospital bill.

I haven’t done my own reporting on any of this, but it’s got me curious, so I want to share what I’ve seen so far — what could be promising, and what’s concerning. And to hear what you think.

But first…

We need your help

Because these newsletters, like our podcast, are all created by hard-working humans, not AI bots.

We love doing this. It’s the best job we could possibly have. We want to do more of it in 2026.

So we’re spending the rest of 2025 raising the money to make that possible.

Now’s the time: please donate to keep us working for you.

Thank you so much!

OK, on to the bots…

Are insurance companies way ahead of us on AI? Maybe not (yet)

After all, as we’ve covered, they’re already using AI to deny claims by the gazillion.

But Claimable’s founder — who spent a couple of years working for the insurance company Anthem — says appeals are different.

“They have to be read by an appropriately qualified human,” he said in the Nocturnists episode.

Could insurance companies juice that process with AI? “Sure, over time,” he says. But: “These companies are often so antiquated on the inside that that’s actually quite a big lift.”

Of course, he’s literally got a stake in selling that answer: Otherwise, how could his company attract customers or investors?

But other recent reporting from the health and science outlet STAT News backs him up. A (paywalled) STAT story describes an AI arms race between hospital billing departments and insurance companies.

And STAT quotes an insurance company CFO admitting that insurers are behind.

For now.

“We’re going to catch up,” he says.

Can I just use an AI assistant myself?

Maybe, in some cases? But as an assistant, not a guru.

That’s my impression after reading the social-media thread by the user who says Claude helped save him $150,000 on a hospital bill.

For one thing, he wasn’t appealing an insurance company decision.

That’s a technical, structured process — and companies offering AI-boosted services say they’ve trained their AI models by feeding them specific, technical data.

This user was fighting the hospital itself over a bill that seemed wildly inflated.

And he did the kind of homework we’ve written about — starting by demanding an itemized bill, with billing codes (which he says the hospital took months to cough up).

And then — with help from Claude — looking up what Medicare would pay.

He says that, with Claude’s help analyzing Medicare rules, he found more than $100,000 in charges that violated those rules.

His next assignment for Claude: Drafting a stern letter to the hospital, “threatening legal action, bad PR, and appearances before legislative committees.”

Here’s his conclusion, about using a tool like Claude, and about fighting with hospitals:

Those last cautions — double-check any facts, treat any AI text as a first draft to edit, not a final copy — sound especially important.

And look: I haven’t done the reporting to verify the claims in this thread.

More than that: I’m as skeptical of AI hype as the next person (unless that person is my colleague Emily, who has me beat there). I especially hate seeing AI slop in my Google results. Recent stories about a possible burst to the AI bubble strike me as long overdue.

But I am cautiously curious about whether there might be something useful here.

If you’ve done some experimenting of your own with any of these tools, you can hit “reply” to this email or hit us up here.

Meanwhile, speaking of caution…

Please don’t just feed AI bots your medical records

Which people seem to be doing, according to a New York Times story out this week. Here’s a gift link so you can read the whole thing.

But here’s the key paragraph, if you’re short on time:

“Inaccurate information is a major concern; some studies have found that people without medical training obtain correct diagnoses from chatbots less than half the time. And uploading sensitive data adds privacy risks in exchange for responses that can seem more personalized.”

That link to the “major concern” about inaccurate info leads to another NYT story (and here’s another gift link): A.I. Is Getting More Powerful, but Its Hallucinations Are Getting Worse.

And the privacy stuff is no joke, since, as The Times notes: health-privacy laws like HIPPA only apply to health-care providers, not tech companies like OpenAI.

Of course, there’s a reason people turn to bots: Actual help from trained, human, health care workers is too tough to access.

As a heading in an earlier NYT story put it: “My doctor is busy, but my chatbot never is.” Ugh.

That’s it for this week! Thanks for reading, and thanks for donating to help us keep doing this work.

… and thanks for the incredible notes some of you have included with those donations. Here’s my favorite from the last week:

You all are doing some of the most important work in these times. I have no doubt that you are saving lives. I’m proud to support monthly and the occasional one-off because I know that your incredible work is making the world a better place. PS — I’ve already had one person tease me about how much I talk about and share your show, because I talk about you All. The. Time.

— Erica, Maryland

Thank you SO much, Erica!!!

And if you’ve donated — or are about to right now — thank you!! Here’s that donation button one more time.

Catch you soon,

— Dan