No Second Chances: How Zus Helps Providers Use AI to Get Risk Capture Right the First Time

CMS is closing the door on retrospective risk adjustment. By 2027, providers won’t get a second chance to fix coding gaps.

That means risk capture becomes a real-time problem. And most organizations aren’t equipped for it. Specifically, 2027 will see CMS eliminate several key chart review practices, such as:

• Diagnoses from unlinked chart review records will be excluded from risk score calculations.
• Providers will no longer be able to revise or optimize coding after billing or beyond tight timelines.

In other words, there are no more “do-overs.” CMS is tasking providers with getting things right in real time. And that shift creates immediate operational risk.

Providers run the risk of losing revenue due to undercoding and incomplete documentation. And since they won’t be able to revise coding issues after billing, clinicians will be under even more pressure to ensure they’re not missing complex issues at the visit. 

Real-time decisions don’t work when the data arrives late. That’s why Zus’ AI-powered prospective risk solution is designed to help providers identify overlooked conditions to enable more accurate risk capture and better patient outcomes.

Here are three ways AI can help stave off that 2027 CMS collision.

1. Expanding the Data Lens Beyond the Patient Chart

A lack of visibility continues to be one of the most limiting factors for providers looking to move to a more proactive approach to patient care.

The main culprit? Traditional healthcare workflows that rely heavily on a single provider’s chart. It works if a patient only sees one provider, but that just isn’t reality for most people.

That’s why aggregating data across multiple providers and historical records is so important. In particular, giving providers access to the full range of a patient’s structured data – such as diagnostic codes like ICD-10 and HCC categories – is critical for identifying risk-adjusting conditions.

This is where most systems break. And it’s where Zus steps in to help providers connect the dots across disparate datasets to surface conditions that may have been previously documented but are no longer visible to the current care team.

For example, AI-sourced HCC opportunities arrive with linked clinical evidence and a clear path to acceptance in the EHR. Zus has a 93% coder acceptance rate on risk suggestions.

2. Unlocking Insights from Unstructured Data

As important as it is, structured data is just one piece of the puzzle. Zus also uses AI to analyze unstructured data, such as clinical notes, to detect evidence of conditions that may not be formally coded. This automates a manual process that would otherwise be extremely time-consuming. 

For example, a patient with diabetes documented in a specialist note but missing from the primary chart could drop out of risk scoring entirely under the new rules. Our system identifies patterns, references, or language indicative of risk-relevant conditions.

The key for clinicians? Explainability. The Zus solution provides traceable evidence, such as highlighted notes or references, so users can evaluate findings.

This dual approach of structured and unstructured analysis is where AI delivers unique value.

3. Turning Risk Insight Into Real-Time Clinical Action

Insight alone doesn’t close risk gaps. Action does. Here’s what the process can look like:

• Step 1: A coder reviews AI-generated insights and flags potential conditions.
• Step 2: A clinician validates these findings during a patient encounter through examination, testing, or direct confirmation.

In this scenario, AI isn’t replacing clinical judgment, but rather augmenting it with a more streamlined operational workflow that can lead to…

• More proactive patient outreach.
• Improved accuracy in diagnosis and documentation.
• Better alignment with value-based care incentives.

Most importantly, this human-in-the-loop model ensures both efficiency and clinical integrity.