Our Take
The dividend exists; the distribution is unresolved and will determine whether AI in health care improves care or just margins.
Why it matters
Health care spending in the US exceeds $4.8 trillion annually. Even modest efficiency gains compound into billions—but those gains will flow to different stakeholders depending on how AI adoption unfolds and who controls the implementation.
Do this week
Health systems: audit your current AI vendor contracts for gain-sharing clauses before renewal cycles close next quarter, so you can negotiate downstream cost offsets before vendors lock pricing.
The AI dividend is shipping, but the bill is unclear
Health care organizations are deploying AI tools to reduce administrative overhead, accelerate diagnostic workflows, and lower the cost of routine clinical decision-making. Vendors report wins in prior authorization automation, radiology interpretation support, and clinical documentation. Providers are seeing measurable time savings and cost reductions in early deployments.
But the gains are not flowing uniformly. Hospitals are capturing efficiency in labor and time; insurers are reducing claims processing costs; vendors are capturing recurring revenue from software licenses and service contracts. Patients are not yet seeing reduced bills or improved access. That asymmetry is the story.
The core tension: who funded the AI infrastructure, who owns the data that trained it, and who gets to set the price for access to it? These questions are largely unresolved in health care, and the answers will determine whether AI becomes a cost lever for providers or a profit lever for platforms.
Distribution shapes whether this is efficiency or rent extraction
AI in health care is a zero-sum game on cost: if insurers pocket the savings from faster processing, they have no incentive to lower premiums. If vendors lock providers into proprietary tools, provider margins compress even as their labor costs fall. If savings accrue only to large health systems with capital to invest in AI infrastructure, smaller providers and rural systems fall further behind.
The health care industry has a history of efficiency gains that do not translate to lower consumer costs. EHR adoption was supposed to reduce administrative burden and improve care coordination; it largely increased documentation workload and shifted costs to providers. The same risk applies to AI if adoption is vendor-led and the business model is extractive rather than aligned.
Early signals suggest the fight is already underway. Providers are demanding that vendor AI tools integrate with existing workflows and provide audit trails showing where savings occur. Insurers are building internal AI capabilities to reduce dependency on vendors. Regulators are beginning to scrutinize AI-driven prior authorization denials. No consensus yet on how gains will be shared.
Document the baseline before deployment
If you are deploying or evaluating AI in health care, measure and lock three metrics before go-live: labor hours saved per process, cost per unit before and after, and time-to-decision for clinical or administrative tasks. Do not rely on vendor estimates or internal projections.
These baselines will determine whether your organization can negotiate fair terms with vendors, defend your adoption ROI against insurance or regulatory review, and identify where margin is genuinely being freed up versus where it is being transferred to a software licensor. The dividend exists; evidence is how you claim your share.