Civilisation’s most underrated technology is not the microchip or the vaccine. It is the checklist. Pilots fly by them, surgeons operate by them, and, as of this year, one of the largest financial verifications in American history runs on them. The checklist is having a moment, and the moment has nine figures attached.
The setting
The story unfolds in Medicare Advantage, the programme through which private insurers cover more than thirty million older Americans. The government pays insurers monthly, adjusted by risk scores built from members’ documented diagnoses, which makes documentation itself worth billions. After years of evidence that documentation had inflated well past plausibility, congressional advisers put the annual excess in the tens of billions of dollars, the government’s audit programme, RADV, scaled up dramatically: roughly two thousand certified coders, quarterly audit cycles, AI-assisted review, and sample error rates extrapolated across entire contracts.
The spring of 2026 supplied the era’s defining numbers. Federal reviews found 81 to 91 percent of certain sampled high-risk diagnosis codes unsupported at three audited plans. A major insurer settled with the Department of Justice for 117.7 million dollars over how its diagnosis records were assembled. And across the industry, a genre of document usually confined to aviation and operating theatres became the hottest artefact in healthcare compliance: the audit-preparation checklist.
Why a list beats a strategy
It sounds bathetic: billion-dollar organisations, facing existential audits, reaching for bullet points. But the checklist’s power in this setting is exactly the power Atul Gawande documented in surgery. When the process is complex, the deadline fixed, and the cost of a skipped step catastrophic, expert judgment alone reliably fails. Not because experts lack knowledge, but because complexity outruns memory under pressure.
An audited insurer faces precisely that shape of problem. When the government selects a contract, the plan must produce, within a five-month window, the medical records behind every diagnosis for a sample of up to two hundred members, records scattered across hundreds of provider offices, spanning care delivered years earlier. Each record must demonstrate that the condition was genuinely monitored, evaluated, assessed, or treated during a real clinical encounter. Miss the window, or submit records that fail the documentation test, and the resulting error rate multiplies across the whole contract.
The industry’s answer is codified in documents like the standard RADV medical record checklist, which decomposes survival into verifiable steps: inventory every audited diagnosis, locate its source encounter, retrieve the note, verify the documentation supports the code, flag the failures for correction, package the evidence to specification, track the deadline. Nothing on the list is intellectually hard. Everything on the list has, somewhere, been forgotten under pressure, at extrapolated cost.
What the best lists contain
Compliance veterans emphasise that the checklist’s real work happens long before any audit letter. The mature version is not a response plan but an operating rhythm: run the government’s test on yourself, quarterly, before the government does.
That self-audit discipline has a revealing signature step: check in both directions. Immature programmes verify that documented diagnoses can be defended. Mature ones also hunt for the reverse failure, recorded conditions that the evidence does not support, and remove them proactively, because prosecutors have made one-directional review the recognised mark of bad faith. The checklist, in other words, encodes ethics as logistics: honesty broken into steps small enough that skipping one becomes visible.
There is also a quietly modern item on the list: know what your AI did. Where software helped identify diagnoses, the record must show the evidence it relied on and the human who confirmed its conclusion. The checklist absorbs the machine age without ceremony, one more step, one more box, same discipline.
The moral of the moment
It is tempting to read the audit era as a technology story, AI reading charts, algorithms checking algorithms. The more accurate reading is older and more human. A system drifted because nobody was systematically checking, and the correction arrived not as genius but as procedure: samples, windows, standards, steps. The checklist is what accountability looks like when it stops being a value and becomes a schedule.
That is the transferable lesson for any field where data has become money, which is gradually all of them. Write down the steps of honesty. Run them on yourself before someone runs them on you. Check both directions. Keep the receipts.
The pilots learned it, then the surgeons. Now the insurers, at 117.7 million dollars a lesson. The humble checklist keeps collecting industries, one crisis at a time, and it is never finished adding lines.
