The most significant problem in healthcare is inefficient workflows. That includes admin work like scheduling appointments and clinical work like visit documentation. So far, I’ve identified 120 inefficiencies in healthcare—and I’ve been writing about a new one weekly.
AI is intended to be our savior from these inefficiencies. If you read my Insights pieces start to finish, you’ll see AI show up in the solution almost every time.
AI will likely make workflows more efficient. The open question is whether it will make them more effective at lowering costs (efficiency vs effectiveness?).
That depends largely on the reimbursement model.
- Fee-for-service: AI improves efficiency, but it also makes it easier to do more billable work. Total costs would likely rise. An AI scribe saves charting time during a typical 15 minute visit and improves billing accuracy, allowing the physician to see more patients per unit time (theoretical; so far, studies show minimal time saving with AI scribes).
- Value-based care: AI improves efficiency, and because reimbursement is less tied to “inputs,” costs could fall. An AI scribe saves charting time during a typical 15 minute visit, allowing the physician to use the full 15 minutes to address patient concerns instead of ending the visit early to see more patients (again, theoretical).
It’s a tricky problem. The authors of a recent NEJM Catalyst article summed it up nicely:
If physician capacity freed up by artificial intelligence scribe services is directed toward actions that can bend the cost curve rather than drive up encounter volumes and billing, longer-term savings are possible through the use of artificial intelligence tools.
Root Cause Analysis: 5 Whys
The 5 Whys process in root cause analysis involves repeatedly asking "Why?" five times to drill down into the root cause of a problem by exploring the cause-and-effect relationships underlying the issue.
The problem: AI may improve clinical workflows while still increasing overall healthcare spending.
- Why?: AI makes clinical work faster by reducing friction in documentation, monitoring, triage, decision support, and administrative tasks.
- Why?: In a fee-for-service system, faster workflows often create more billable capacity rather than lower total costs.
- Why?: More capacity can lead to more visits, tests, referrals, prescriptions, prior authorization activity, and downstream care cascades.
- Why?: Even when AI lowers the labor cost of delivering care, consolidated health systems, payers, and vendors may keep those productivity gains as margin instead of passing savings to patients or purchasers.
-
Why (root cause)?: AI is being layered onto a healthcare system whose financial incentives still reward volume, complexity, and revenue capture more than lower total cost of care.
Impact Analysis
Impact analysis is the assessment of the potential consequences and effects that changes in one part of a system may have on other parts of the system or the whole.
- Patient: Patients may get better access to care through faster appointments, more digital touchpoints, remote monitoring, AI triage, and earlier detection of disease, but they may also face more downstream testing, referrals, prescriptions, and out-of-pocket costs if AI expands the amount of care being delivered rather than replacing higher-cost care.
- Clinician or Provider: Clinicians may spend less time on documentation, chart review, inbox work, and administrative tasks, but that saved time may be redirected into higher patient volumes, more clinical throughput, and more responsibility to review AI-generated recommendations, escalations, and incidental findings.
- System: AI may improve efficiency and margins for health systems, payers, and vendors, but unless payment incentives shift toward value-based care, the system is likely to convert productivity gains into more billable services, higher utilization, and higher total spending rather than lower costs for patients and purchasers.
The Thought Experiment
The article I want to highlight is from Bob Kocher, M.D., Brian Zhao, and Erin Duffy, Ph.D., M.P.H. in NEJM Catalyst. Their main argument is straightforward:
AI will probably improve parts of healthcare, but that does not automatically mean it will lower healthcare spending.They walk through several areas where AI could change costs:
- Drug development: AI may help companies identify targets faster, recruit patients into trials more efficiently, and bring new therapies to market sooner. That could create long-term value, especially for chronic diseases where better treatment prevents expensive complications. But more new drugs also means more high-cost drugs entering the market, and in the U.S., that usually means higher pharmaceutical spending.
- Clinical access: AI scribes, remote patient monitoring, chronic care management, direct-to-consumer AI care, and clinical decision support could all make care easier to deliver. Patients may be seen sooner. Physicians may spend less time documenting. High-risk patients may be monitored more closely. If AI creates more visits, more testing, more referrals, more prescriptions, and more incidental findings, total spending can rise quickly.
- Administrative work: AI may reduce labor costs in scheduling, coding, billing, claims processing, prior authorization, and utilization management. That feels like the clearest path to savings (and it’s largely where I focus in my Insights articles). Kocher, Zhao, and Duffy make an important point, though: lower administrative costs do not automatically translate into lower prices for patients. In concentrated hospital and insurance markets, those savings may become margin (a point I’ve missed in prior articles).
The core idea is that AI’s cost impact depends less on the technology itself and more on the business model around it. Under fee-for-service, AI can make it easier to do more billable work. Under value-based care, where the incentives are different, AI can (theoretically) target high-risk patients, prevent avoidable utilization, reduce unnecessary services, and make clinicians more effective within a fixed budget.
AI can make healthcare more efficient. Inside the wrong reimbursement model, efficiency often converts into more healthcare activity. The technology may be new, but the financial incentives are very familiar.
As Mr. Deming once said: Every system is perfectly designed to get the results it does.
In summary, AI alone will not bend the cost curve. If we want AI to lower spending, the reimbursement model has to reward lower total cost of care (not just faster throughput, higher volume, and better revenue capture).
Read the original post here.
