The Centers for Medicare & Medicaid Services is putting real money behind the idea that artificial intelligence can catch bad claims before they’re paid, not just after. On September 28, 2026, Nextgov/FCW reported that CMS Deputy Administrator and Chief Operating Officer Kimberly Brandt described a modernization push built around machine learning, real-time claims validation, and a growing set of vendor contracts aimed squarely at documentation-driven improper payments. For coding and compliance teams, this is less a fraud story than a signal about what payers will expect from documentation going forward.
What CMS is actually building
The centerpiece is the agency’s Fraud Defense Operations Center, launched in March 2025, which Brandt says has stopped nearly $2.5 billion in potentially fraudulent Medicare payments and taken action against roughly 800 providers. That center is now being paired with a modernized Medicare claims processing system designed for real-time payment capabilities, rather than the retrospective pay-and-chase model CMS has relied on for decades.
The contracts behind the shift
Two contracts anchor the buildout: HealthEdge, with a $2.5 million initial obligation and a $1.1 billion ceiling, and Peraton, with a $9.1 million initial obligation and an $825 million ceiling. Tegria has been brought on for implementation support. Together they point to heat-mapping algorithms that flag frequently billed items and AI-powered claims analysis meant to surface documentation patterns that warrant human review — before payment goes out, not after an audit letter arrives.
Why documentation is the real target
Brandt was direct about where the savings are expected to come from: “If we can reduce that documentation improper payment, that would drop our entire improper payment rate down.” That’s a notable framing. CMS isn’t primarily describing this as a hunt for bad actors — it’s describing a system built to catch the much larger, much more mundane problem of claims that get denied or clawed back because the documentation didn’t support the code billed, even when the underlying care was appropriate.
That distinction matters for how coding teams should read this. A heat-mapping algorithm doesn’t know intent. It flags patterns: a code paired with documentation that looks thin relative to similar claims, a specialty billing at rates that diverge from its peers, a note that’s missing the specific elements a payer’s coverage policy requires. Real-time review means those patterns get caught closer to the point of submission, which raises the cost of documentation gaps that used to slip through and only surface months later in a RADV or CERT-style audit.
It’s also worth noting what this isn’t. CMS isn’t proposing new coding rules or a new set of documentation requirements — the underlying guidance for ICD-10-CM, CPT, and payer coverage policy stays the same. What changes is enforcement speed and consistency. A documentation gap that once had a reasonable chance of going unnoticed for a year or more, buried in claims volume, is now a candidate for same-cycle detection. That shrinks the margin for error that many revenue cycle teams have, whether they realize it or not, been quietly relying on.
What this changes for coders and compliance teams
None of this changes coding guidelines. It changes the tolerance for the gap between what a chart actually supports and what gets billed. A few practical implications stand out:
- Documentation-support gaps that once took months to surface in an audit may now get flagged at or near submission, shrinking the window to fix or appeal.
- Coding accuracy and clinical documentation integrity become more closely linked — a code that’s technically correct but thinly supported is now a bigger liability than before.
- High-volume, high-variance code sets (the kind CMS’s heat maps are built to spot) deserve a fresh look at documentation templates and query workflows.
- Compliance teams should expect payers to follow CMS’s lead — commercial plans have historically mirrored Medicare program-integrity tooling within a year or two of its rollout.
The upstream fix CMS isn’t building
CMS’s approach is fundamentally a downstream catch: better detection of problems after a claim is coded and submitted. It doesn’t solve the upstream problem of coders and CDI staff not having the tools to verify, before submission, that documentation actually supports a given code against current payer requirements. That’s the gap an automated coding and documentation-verification workflow is built to close — checking codes against source documentation and payer policy before a claim ever reaches a CMS heat map, rather than reacting after the fact.
Commercial payers watching this rollout will likely draw the same conclusion CMS did: catching documentation mismatches earlier is cheaper than paying and clawing back later. Coding teams that wait for that shift to show up in payer edits, rather than building the verification step into their own workflow now, will be reacting to someone else’s timeline instead of setting their own.
The bottom line
CMS’s AI buildout is a compliance story before it’s a technology story. The dollar figures are large, but the operational message for coding teams is simple: documentation quality is about to be checked faster, more systematically, and closer to the point of submission than it has been before. Teams that treat documentation-support review as a pre-submission discipline, rather than a post-denial cleanup task, will be the ones least exposed when CMS’s new systems go live.
This is exactly the gap Medikode’s automated medical coding platform is built to close — verifying that documentation supports every code before a claim goes out the door, not after CMS’s algorithms flag it.