VBC is the real AI ROI question. FFS AI is a cost center.
AI investments under fee-for-service reimbursement cap out at administrative efficiency. Only value-based care contract structures convert AI into compounding margin. The contracting model determines the AI ROI, not the model.
Every healthcare AI ROI conversation in 2026 starts with the same three numbers: hours saved on documentation, percentage reduction in prior-auth processing time, percentage decrease in claim denials. The numbers are real. They are also small. A health system running AI under a fee-for-service contract structure ends up with a tighter administrative engine and the same revenue. The provider is more efficient at producing the same volume of billable encounters. The savings flow to the operating budget; they do not compound into strategic margin.
That is the structural ceiling on FFS healthcare AI: under fee-for-service, AI is a cost-reduction tool. Under value-based care, AI is a margin-expansion tool. The contracting model — not the AI — determines whether the investment compounds. Most provider organizations have not internalized the implication: their AI ROI math is determined by their VBC contract penetration, and provider organizations underweight in VBC are buying a smaller version of AI than the technology actually offers.
Why the contract structure is the ceiling.
FFS AI optimizes the cost of producing a billable encounter. That is structurally bounded.
Under fee-for-service, every dollar of clinical work has to flow through a billable encounter to produce revenue. AI that makes encounters cheaper to produce — ambient scribes, automated coding, prior-auth optimization, denial-management — reduces the cost of revenue but not the revenue itself. The savings cap out at roughly the cost of the administrative burden, which is real but finite. The operator who optimizes for FFS AI ROI hits the ceiling within 18–24 months and discovers the next dollar of investment has nowhere to go.
VBC AI optimizes the outcomes inside a fixed-revenue envelope. That compounds without limit.
Under value-based care contract structures (capitation, bundled payments, shared savings, full risk), the provider is paid a defined dollar amount per attributed life or per care episode. Revenue is fixed; the operator’s margin is the difference between the contract revenue and the cost of the care delivered. AI that improves clinical outcomes reduces avoidable utilization (ED visits, readmissions, complications) which is pure margin lift. Every dollar of AI investment that reduces utilization compounds directly into margin, with no administrative ceiling.
The same AI deployment produces different ROI under different contracts.
An ambient scribe deployed at an FFS practice produces hours-saved as the ROI metric. The same scribe deployed at an ACO with shared-savings contracts produces hours-saved plus documentation quality improvement that translates to better risk-adjustment, better quality-measure capture, and better patient outcomes — all of which compound into shared-savings revenue. The technology is identical. The ROI is structurally larger.
Provider AI strategy should be sequenced after VBC strategy, not before it.
The provider organization that starts its AI strategy with a platform decision and works backward to use cases will misallocate. The right sequence is: name the VBC contracts you are signed into or signing into; identify the metrics those contracts grade you on; then deploy AI against those metrics. The AI investment compounds with the contract structure. The provider that gets this sequence backwards buys efficient FFS production and wonders why the strategic upside is missing.
The strongest argument against this position.
The strongest counter is that VBC penetration in US healthcare is uneven and many provider organizations realistically cannot move their book of business into VBC quickly enough for this argument to be operational. This is true. The piece is not arguing that every provider should immediately convert to full-risk capitation. It is arguing that the AI ROI conversation has to be paired with the VBC strategy conversation. A provider with a 5% VBC book and a 95% FFS book should expect 5% of the AI ROI math to compound and 95% to cap out at administrative efficiency. That is fine — but it should be planned for, not discovered three years in when the AI program plateaus.
Three things to do this quarter.
01 · Build your AI ROI model with explicit FFS and VBC line items. What share of expected return depends on FFS administrative efficiency? What share depends on VBC outcomes? If the model doesn’t separate the two, it is overstating one or the other.
02 · Sequence the VBC strategy decision before the AI platform decision. AI is a force multiplier on the contract structure. Pick the contracts first.
03 · For each AI use case, name which VBC contract metric it improves. If the use case doesn’t map to a VBC metric, classify it as FFS-only and budget it as administrative cost reduction, not strategic growth.
SALT’s position-review rhythm grades published positions against subsequent reality.