BJC Healthcare's CFO said it out loud at HFMA 2026: "Bot versus bot — provider revenue cycle bot against the denial insurance bot — is the fundamental problem." Scott Hawig wasn't describing a future risk. He was describing what's happening right now in thousands of hospitals across the country — and why billions in AI investment are producing stalemate instead of results.
The healthcare industry is spending more on AI than ever. Capital is flowing, vendors are multiplying, and every revenue cycle conference features the word "agentic" on every other slide. But the returns aren't matching the investment. HFMA 2026 made the reason unmistakably clear: most hospitals are bolting sophisticated AI onto decades-old infrastructure, and the result is a more expensive way of staying stuck.
The Bot-vs-Bot Stalemate, Explained
Here's the dynamic that Hawig is describing. Payers deploy AI to adjudicate claims faster, flag utilization outliers, and auto-generate denials. Hospitals respond by deploying AI to scrub claims, generate appeals, and automate prior authorization. Both sides are now automating at machine speed.
The problem? Neither side is fixing the underlying data. Both AI systems are ingesting the same fragmented, inconsistent, legacy-system-generated information and simply processing it faster. The payer bot denies faster. The provider bot appeals faster. The volume increases. The net outcome doesn't.
This is the revenue cycle equivalent of an arms race where both sides keep upgrading weapons but fighting over the same trench. Speed increases. Resolution doesn't. And the organizations paying for both sets of AI tools absorb the cost of the stalemate.
The Data Strategy Gap: What HFMA 2026 Leaders Actually Said
The most quoted line at HFMA 2026 wasn't about a new product launch or a merger announcement. It came from Seema Verma — now at Oracle Health, formerly the CMS administrator who oversaw interoperability rules for the entire U.S. healthcare system:
"You can't have an AI strategy without having a data strategy."
Verma argued that the infrastructure most providers work with is too antiquated for AI to deliver on its promises. The data exists — but it's trapped in systems that were never designed to talk to each other, let alone feed real-time AI models.
HCA Healthcare CFO Mike Marks reinforced the point from the operator's perspective: "The amount of legacy systems that we're all dealing with is really getting in the way of transformation." Marks prioritized clinical systems first, then operations and administration — acknowledging that transformation isn't a single initiative but a sequenced infrastructure overhaul.
Together, these leaders painted a picture that every hospital CFO needs to internalize: writing huge checks for AI without fixing the data foundation is a more expensive way of staying stuck.
Why Legacy Systems Are the Real Bottleneck
Most hospital revenue cycles run on infrastructure built 20 to 40 years ago. These systems were designed for batch processing, manual workflows, and paper-first communication. They store data in proprietary formats, use interfaces that resist modern integration, and can't support the throughput AI models require.
The practical consequences are severe:
- Fragmented patient records — Eligibility data lives in one system, clinical documentation in another, coding in a third. AI agents that only see one piece can't prevent downstream failures.
- Inconsistent coding standards — When five departments encode the same procedure differently, AI trained on that data inherits and amplifies the inconsistency.
- No real-time data exchange — Batch-processed files mean that by the time an AI agent identifies a problem, the claim has already been submitted, denied, and aged.
- Siloed department systems — Scheduling, registration, coding, billing, collections, and patient communication each operate in isolation. An AI agent optimizing one silo can actively hurt another.
This is why hospital AI agents that bolt onto individual workflows produce marginal gains at best and new problems at worst. The AI isn't broken — the data it operates on is.
The Outsourcing Trap: Same Problem, Different Wrapper
HFMA 2026 also highlighted a parallel pattern: major health systems outsourcing their revenue cycles rather than building in-house AI capability. IU Health entered an RCM partnership with Ensemble Health Partners in June 2026, joining a growing list of systems that have decided the internal transformation is too hard.
But outsourcing doesn't solve the data problem — it just pays someone else to manage the same broken infrastructure. The legacy systems don't disappear. The fragmented data doesn't unify. And the hospital loses direct control over the one thing that determines AI effectiveness: the quality and architecture of its own data.
The organizations that will win this transition aren't the ones who outsource the problem or the ones who buy the most AI tools. They're the ones who fix the foundation first, then deploy AI that can actually leverage it.
What a Data-First AI Strategy Actually Looks Like
Moving from the bot-vs-bot stalemate to genuine AI-driven revenue improvement requires a sequenced approach. Based on the consensus emerging from HFMA 2026 — and the operational reality of running AI across complex health systems — here's what that sequence looks like:
1. Unify the Patient Financial Record
Before deploying any AI, establish a single longitudinal view of each patient's financial journey — from eligibility verification through final payment. This means connecting scheduling, registration, insurance verification, coding, billing, and collections data into a shared context layer that every AI agent can access.
2. Standardize Data Pipelines
Eliminate the batch-processing bottleneck. Real-time data pipelines ensure that eligibility changes, clinical documentation updates, and payer rule modifications flow to every relevant AI agent simultaneously. When a patient's coverage changes at 2 PM, the claim submission agent should know by 2:01 — not the next morning.
3. Deploy Agentic AI Across the Full Cycle
Once the data foundation is sound, deploy AI agents that operate across the entire revenue cycle with shared context — not as disconnected point solutions. An AI denial management agent that knows what happened at eligibility, coding, and submission can prevent denials before they occur. A prior authorization agent that understands the patient's full clinical and financial picture can assemble stronger requests from the start.
4. Build Feedback Loops That Learn
The final piece — and the one most "AI strategies" skip — is building systems that learn from every claim, every denial, every appeal, and every payment. Leading health systems are moving beyond "tool sprawl" to build revenue cycles that learn from every encounter. This isn't about running reports quarterly. It's about AI agents that update their models continuously based on real outcomes.
The HFMA 2026 Optimism Signal
Despite the sobering reality of the bot-vs-bot stalemate, HFMA 2026 wasn't all doom. Post-conference synthesis from Chief Healthcare Executive reported "surprising optimism" among healthcare leaders — with multiple executives expressing enthusiasm for AI's potential to improve operations, help overworked staff, and drive down costs.
The optimism isn't naive. It's rooted in a specific insight: the opportunity to find revenue "that is still recoverable and addressing it before it ages, is reduced, or is written off." In other words, the leaders who are optimistic aren't expecting AI to create new revenue from nothing. They're expecting AI to stop the hemorrhaging — to catch the claims that shouldn't have been denied, flag the authorizations that are about to expire, and identify the revenue that's sitting in aging A/R buckets.
But that optimism comes with a condition: the data has to be right first.
The CMS Regulatory Tailwind
The data strategy imperative isn't just an operational preference — regulators are pushing in the same direction. The AMA's updated prior authorization game plan requires same-specialty physician reviewer oversight and mandates that automated systems cannot supplant individualized clinical judgment. CMS requires Medicare Advantage, Medicaid, and ACA plans to respond to urgent prior authorizations within 72 hours, standard requests within 7 days, and implement FHIR APIs by 2027.
These rules create a paradox for the bot-vs-bot dynamic: payers must divulge denial reasons and report prior authorization metrics, which means provider-side AI that's built on a clean data foundation can exploit that transparency to build stronger claims and appeals. The hospitals that have their data in order will be the ones best positioned to use payer transparency as a competitive weapon.
Where BAM AI Fits: Data Foundation + Agentic Revenue Cycle
BAM AI's approach was built for exactly this moment. Instead of selling another point solution that bolts onto legacy systems, we deploy agentic AI that operates across the full revenue cycle — from eligibility verification through denial management and appeals — with shared patient context at every stage.
The difference from the bot-vs-bot dynamic:
- Unified data layer — Every AI agent accesses the same patient financial record. Eligibility failures don't cascade silently into downstream denials.
- Prevention over reaction — Instead of generating faster appeals after denials, our agents prevent denials before submission by catching issues at the source.
- Continuous learning — Every claim outcome feeds back into every agent's decision model. The system gets smarter with every encounter, not just faster.
- No legacy system dependency — BAM AI integrates with your existing EHR and PM systems without requiring a full infrastructure overhaul. The unified data layer sits alongside your current stack, not instead of it.
The bot-vs-bot stalemate exists because most AI tools are optimized for speed, not accuracy. BAM AI is optimized for outcomes — which starts with the data.