What if AI competed on healthcare outcomes?

AI systems competing on health outcomes would shift incentives away from selling software volume and towards verifiable improvements in morbidity, mortality, cost and equity, but it would require new contracting, measurement and governance to avoid gaming and exacerbating disparities.​

Core idea

If AI vendors were paid primarily for outcomes (e.g., fewer readmissions, lower HbA1c, earlier cancer stage at diagnosis), they would be incentivized to build tools that measurably improve population health rather than just workflow or billing metrics. This is essentially extending value-based care logic from providers to the AI supply chain, aligning payment with validated quality metrics and total cost of care.​

How AI could “compete” on outcomes

1) Tie payment to pre-agreed clinical metrics

Contracts can reward AI only when specific endpoints move, such as reduced heart failure admissions, better diabetes control, or improved screening uptake.​
Performance-based linear and “dynamic” models already being piloted share upside with vendors when outcomes improve, and claw back fees when they do not.​

2) Use value-based reimbursement structures

Proposed frameworks include per-patient or per-episode fees contingent on achieving quality standards, time-limited add-on payments, and bonuses for demonstrable outcome gains instead of per-scan or per-click fees.​

For generalist medical AI, tiered reimbursement aligned with level of autonomy (assistive → augmentative → autonomous) could be coupled to outcome benchmarks at each tier.​

3) Outcome-focused use cases

In value-based care programs, AI already supports risk prediction, early intervention and coordinated care, with reported 30–45% cost savings in high-cost episodes when outcomes improve.​

AI-driven personalized prevention and screening programs have increased screening rates and early cancer detection, illustrating how outcome-linked incentives for insurers and vendors can form a virtuous cycle.​

Benefits of outcome-based AI competition

a) Better alignment of incentives

Payers can use reimbursement as a “gatekeeper” to prioritize AI that demonstrably improves workflows, access and outcomes rather than throughput or high-margin services.​
When life and health insurers share in financial gains from healthier members, they are natural funders of AI that drives sustained behaviour change and prevention.​

b) Stronger ROI discipline

Performance-based pricing addresses buyer skepticism about paying high prices for AI without clear, realized value.​

Outcomes-based models support investment into prevention, chronic disease control and post-acute optimization that are often under-incentivized in fee-for-service.​

c) Potential to reduce inequities

Properly designed AI can improve early detection, adherence and post-treatment monitoring in underserved populations, closing outcome gaps if metrics are stratified by race, gender and geography.​

Some policy proposals explicitly suggest rewarding AI that improves equity and penalizing tools that increase bias.​

Risks and failure modes

i) Measurement and attribution challenges

Outcomes such as mortality, readmissions or A1c can take time to move and are influenced by many factors, making causal attribution to a single AI tool nontrivial.​

Over-reliance on narrow metrics can drive gaming, such as avoiding high-risk patients, over-documenting, or optimizing for surrogate endpoints that do not translate into genuine health gains.​

ii) Equity and bias concerns

There is documented concern that AI can encode and amplify existing biases, worsening disparities unless datasets, objectives and incentives explicitly correct for this.​

If vendors are rewarded on average performance without equity constraints, they may focus on “easy” populations and neglect complex or disadvantaged groups.​

ii) Market structure and lock-in

Sophisticated outcome-based contracts could favour large incumbents who can carry risk, build data-sharing infrastructure, and tolerate delayed payments.​

Payer and provider dependence on a small number of high-performing platforms may reduce contestability unless interoperability and switching are also rewarded.​

What it would take to make this work

a) Robust real-world evaluation infrastructure

Health systems need standardized evaluation pipelines, including prospective, multi-site studies and RWD-based monitoring, to validate that AI actually improves outcomes before and after deployment.​

National programs (e.g., AI in Health and Care evaluations) already highlight the need for careful measurement, bias assessment and ongoing post-deployment surveillance.​

b) Contract and policy innovation

Regulators and payers are exploring targeted outcome incentives, transitional add-on payments, advance market commitments and reimbursement standards that explicitly include interoperability and bias mitigation.​

Performance-based pricing schemas in which fees escalate with realized outcome improvements (and fall with underperformance) are emerging templates for AI vendors.​

c) Governance and transparency

Responsible AI frameworks stress explainability, continuous monitoring, multidisciplinary oversight and clear allocation of liability between clinicians, institutions and vendors.​

To avoid perverse incentives, contracts will likely need guardrails around patient selection, minimum quality thresholds, and equity-adjusted metrics.​

For someone operating in healthtech M&A, the implication is that a significant portion of future AI value will be priced off validated outcome improvements and risk-sharing capacity rather than mere model performance, which will directly influence business models, revenue visibility and valuation approaches.

To discuss how Nelson Advisors can help your HealthTech, MedTech, Health AI or Digital Health company, please email [email protected]