At the HFMA Annual Conference, health system CFOs repeatedly highlighted a shared challenge: strong performance within individual functions doesn’t reliably translate into stable end-to-end outcomes. Each team optimizes its own piece. CDI, coding, utilization review, and billing all hit their targets, but the same clinical encounter gets reinterpreted differently at every handoff. With no shared logic, different rulesets and decision frameworks are applied at each stage, creating variation and rework that affect downstream financial outcomes.
That’s where performance erodes: between functions rather than within them.
This disconnect is driving the surge in CFO interest in mid-cycle AI. The goal has moved past fixing isolated workflows toward keeping decisions consistent as they travel through the cycle, so what gets decided in one process holds up in the next.
Here’s what CFOs are saying.
Why the Mid-Cycle is Gaining CFO Attention
The mid-cycle sits between clinical care and financial performance, where reimbursement is defined. Decisions made here shape downstream outcomes like denials, payment variance, and A/R, while also feeding back into upstream processes like documentation quality and authorization workflows. This is where redundancy becomes the most visible and costly. Because the same clinical encounter is evaluated by multiple teams using different rulesets, each re-interpretation introduces variation and delay as information moves between CDI, coding, utilization review, and billing.
That’s why CFOs are turning to mid-cycle AI applications to minimize rework and align handoffs. AI deployed here learns from payer behavior, clinical context, and historical resolution patterns, then applies shared logic across processes instead of as isolated, function-specific rules. This approach shifts the revenue cycle from fragmented decision-making to an integrated system that learns and adapts.
We’ve reached a point where the real constraint in revenue cycle is how consistently information, decisions, and context move upstream into patient access or downstream into reimbursement. That’s where the financial impact is either realized or lost.
- CFO, Midwest academic health system
Mid-Cycle AI Use Cases Gaining Traction
Conference attendees made clear that the most effective mid-cycle AI applications are not broad automation platforms, but targeted decision-support tools embedded directly into existing clinical and financial workflows.
What’s gaining the most traction are the use cases that reduce rework at key handoffs, improve the completeness of clinical information before downstream processing, and help teams operate from a shared view of the same encounter. Value is showing up fastest where AI is tightly scoped, uses the human-in-the-loop approach, and connects to measurable financial outcomes.
Four areas were repeatedly cited as where AI is already delivering results in production environments:
1. Clinical Documentation Integrity (CDI)
AI in CDI is reducing documentation-driven denials, increasing first-pass yield, lowering coding rework, and decreasing cost per case by resolving gaps earlier in the revenue cycle, often before coding or denial activity occurs. AI surfaces risk, missing specificity, and inconsistencies, while clinicians and CDI specialists determine clinical validity and documentation actionability.
The AI Making a Difference:
- LLMs for documentation gaps and leakage prevention
- NLP and embeddings for denial, underpayment, and case mix pattern linkage
- RAG systems for policy- and outcome-aligned reimbursement decisions
More of the documentation issues that used to show up as coding delays or denials are being resolved earlier in CDI, which is improving first-pass yield.
- CFO, Southern academic health system
2. Medical Coding
Coding directly drives reimbursement accuracy, DRG assignment, denial risk, and case mix stability, making it one of the most financially sensitive mid-cycle functions. AI adoption is improving coding speed, reducing denial rates, increasing throughput, and lowering cost-to-collect by reducing manual work.
Most Mature AI Solutions:
- Learning models trained on historical coding decisions and outcomes
- NLP for extracting diagnoses, procedures, and clinical entities from unstructured notes
- Rule-based overlays for payer-specific coding constraints and local policy variation
The coders are still making the final calls, but a larger share of their time is spent validating recommendations versus assembling the full clinical picture from scratch. The human-in-the-loop model has been a game-changer for efficiency and accuracy.
- CFO, Midwest hospital
3. Denial Management
AI insights generated through denial management can be applied upstream to improve documentation, coding, and utilization review practices, helping increase first-pass payment yield and reduce margin leakage.
Top AI Investments:
- AI models that detect denial patterns across payers and care settings
- Predictive models for denial risk and preventable revenue loss
- Systems linking denial drivers to upstream clinical and coding decisions
We’re using denial data more strategically now with AI. It’s feeding back into upstream documentation and coding decisions instead of just sitting in reporting dashboards.
- CFO, Southern health system
4. Utilization Review
Utilization review has a direct impact on authorization success rates, DNFB days, reimbursement timing, cash acceleration, and revenue predictability. AI is improving clinical completeness before payer submission, reducing preventable denials and accelerating reimbursement cycles.
Impactful AI:
- Payer rules engines encoding medical necessity criteria and authorization requirements
- Supervised learning models trained on historical authorization outcomes
- Queue optimization models for prioritizing high-risk or high-variability cases
AI is surfacing missing details earlier in the process. Fewer cases are entering the denial pipeline because utilization review is more complete before submission.
- CRO, Northeast hospital
Mid-Cycle AI and the Next Era of Revenue Cycle
What’s becoming clear from CFO discussions is that revenue cycle performance is no longer constrained by the effectiveness of individual functions but by how reliably information moves across them. Mid-cycle sits at the center of that dynamic, where data is fluid enough to influence downstream financial outcomes, yet structured enough to affect what happens upstream.
For CFOs, this reframes what mid-cycle AI is actually buying. Not faster individual functions, but a system where one team’s output actively shapes the others.
