Our Take
Good technology alone has never closed the gap between AI ambition and clinical reality; health systems need to rebuild their revenue cycle and operational workflows first.
Why it matters
Revenue cycle management (RCM) remains one of healthcare's largest cost drains, and AI vendors are flooding the market with claims that algorithms solve the problem. They don't. Health systems need governance, process redesign, and clinician buy-in before any model matters.
Do this week
RCM leads: audit your current RCM vendor's deployment success rate (claims of accuracy mean nothing without post-implementation adoption data) before renewing or upgrading licenses.
HIMSS convenes healthcare executives on AI implementation reality
HIMSS Media is hosting two regional summits focused on moving health systems from AI pilots to operational deployment. The AI in Healthcare Forum runs June 25-26 in Boston, bringing together clinicians, executives, technologists, and researchers for what the organizers describe as "two immersive days focused on the real-world application of AI in health and care." A second gathering, HIMSS26 APAC, follows August 23-25 in Singapore, aimed at health system leaders, government voices, and technology partners across the region.
The timing reflects an industry-wide tension. Healthcare vendors have shipped hundreds of AI-enabled RCM tools over the past 18 months, each claiming to reduce claim denials, accelerate collections, or predict patient payment risk. Yet most health systems report that these tools sit idle or deliver far lower returns than promised in pilots. The gap between vendor claims and operational reality has become acute enough that a major industry conference is now dedicating entire sessions to the mechanics of actual deployment.
Process design beats model quality in healthcare AI adoption
Revenue cycle management is healthcare's largest unresolved operational cost. A single denied claim can cascade into months of rework, staff overtime, and revenue leakage. AI vendors have positioned RCM automation as the solution: use machine learning to flag high-risk claims before submission, predict patient ability to pay, or automate denial appeals.
The problem is not the algorithms. It is that deploying them requires health systems to redesign workflows that billing teams have optimized over decades. Billing staff become defensive when told an AI model will flag their work as error-prone. Clinicians resist changes to charge capture workflows. Finance teams struggle to trust model predictions without transparent reasoning. None of these barriers are technical.
HIMSS's framing of "evidence-based action" signals recognition that vendor benchmarks (accuracy on historical data, reduction in denied claims in controlled pilots) do not predict operational success. Health systems need implementation models, governance structures, and staff retraining protocols. The algorithms are table stakes; the operating model is the actual product.
What health system executives should demand from AI vendors
If you are evaluating an AI-enabled RCM tool, stop requesting accuracy metrics from the vendor's test set. Instead, ask for post-deployment adoption rates and revenue impact from three comparable health systems (comparable in size, payer mix, and EHR platform). Ask how the vendor trains billing staff, who owns process redesign, and how long deployment typically takes from go-live to breakeven.
Most vendors will not have clean answers to these questions, which is itself the answer. A tool that requires your organization to rebuild your billing workflows is not a software purchase; it is a transformation project. Price and staff it accordingly. And attend sessions at Boston or Singapore that focus on the non-technical barriers: governance, change management, clinician alignment, and evidence collection. That is where the real work lives.