Frequently Asked Questions

Autonomous medical coding is the next generation of coding automation. Using advanced artificial intelligence, it reads clinical reports and assigns the correct CPT, ICD-10-CM, and HCPCS codes automatically without human intervention.

Unlike older coding technology, the AI engine understands the context of the entire report, not just individual words, which allows it to code with a high degree of accuracy.

It leverages multiple AI methods:

• Machine Learning (ML): Learns from structured historical data.
• Deep Learning (DL): Uses multilayered neural networks to recognize complex patterns.
• Transformer Models: A form of predictive machine learning, the AI interprets entire clinical reports in context rather than relying on keywords.

 

By combining these methods, the system mimics human reasoning, enabling it to process unstructured clinical data, interpret reports, and assign accurate codes within seconds without human intervention.

Maverick trains and fine-tunes these models using medical coding data to create a Masked Language Model (MLM). The MLM is designed to perform medical coding with the highest accuracy standards, ensuring precise and reliable outcomes.

Key capabilities of the Maverick proprietary AI engine include:

• High coding accuracy that supports 85%+ of claims going direct-to-bill (DTB) without manual/human review.
• Ability to process both structured and unstructured data.
• Continuous learning from new documentation and coding patterns.
• Automation that eliminates the need for human intervention while still maintaining compliance.

Autonomous coding is not simply an advanced form of CAC; it represents an entirely different class of technology. While CAC offers suggested codes for human coders to review and finalize, autonomous coding performs the entire coding process independently – without human intervention.

Unlike CAC, which relies on rigid rule sets and structured documentation, autonomous coding leverages advanced machine learning, deep learning, and transformer models trained on vast historical datasets. It involves our AI model analyzing large volumes of unstructured clinical data and understanding context. The models capture nuanced variations in provider documentation and coding styles, enabling more consistent and accurate code assignment.

As coders code exceptions (those cases the model could not code with high confidence to send direct to bill), the system learns from their feedback, continuously improving its precision. The result is a transformative leap in performance – autonomous coding achieves high direct to bill (DTB) rate, a benchmark that CAC has consistently struggled to achieve.

Autonomous coding transforms the revenue cycle into a faster, smarter, and more cost-effective operation. By achieving high DTB rates and reducing manual intervention, organizations benefit from stronger compliance, improved financial performance, and enhanced operational resilience. The following represent both immediate and long-term returns:

• Faster coding – returns codes in seconds, running 24/7
• Eliminates coding backlogs
• Accelerates billing cycles and cash flow
• Ensures coding consistency and compliance
• Minimizes denials and appeals
• Decreases reliance on outsourced coding (offshore or onshore)
• Scales to meet growing procedural volumes without proportional staffing increases

• Codes radiology reports in seconds with high accuracy.
• Sends most claims directly to billing without human review. (our customers have experienced up to 93% DTB)
• Routes exceptions (e.g., low confidence cases, missing documentation, highly complex IR cases, etc.) to coders or radiologists for completion.
• Minimizes coding delays, accelerates revenue capture, and enables scalable growth – all while reducing operational costs.

Maverick trained the base model using 10s of millions of radiology reports. The foundation of high DTB performance lies in how the base-model is further trained and optimized for each client. Our approach uses the past two years of your historical data, allowing our proprietary model to learn how your coders consistently apply CPT®, HCPCS, and ICD-10-CM codes. By adapting to your documentation patterns, coding practices, and case complexity, the model is tuned specifically to your environment rather than relying on generalized coding rules. Advanced transformer-based deep learning enables it to recognize subtle clinical nuances that traditional coding tools often miss, laying the groundwork for an 85%+ direct-to-bill rate.

With a continuous live feed, reports are coded and sent to billing in as little as two seconds—24 hours a day, 7 days a week. For organizations that prefer batch processing, we also support this method. As each batch is processed, encounters begin flowing into the platform and are sent direct-to-bill (DTB) without delay.

Autonomous coding solutions send exceptions to human review for the following reasons:

• Low-confidence predictions
• Missing or insufficient documentation
• Ambiguous or conflicting language
• Procedures that may not meet medical necessity (LCD or NCD edits)
• National Correct Coding Initiative (NCCI) edits 
• Complex cases (e.g., interventional radiology)

Yes. The model includes built-in explainability features that show which portions of the clinical documentation supported each code assignment. This allows users to trace the reasoning behind every decision, making it easy to validate accuracy, support compliance, and build trust in the system. Additionally, all encounters sent direct-to-bill (DTB) are available for full review at any time, giving your team complete visibility into coding outcomes.

Prior to implementation, multiple rounds of coding validation take place where we evaluate code mismatches between the Maverick model and historical coding data. While this isn’t a formal audit, it functions as a quality check. If we identify errors in the historical coding or gaps in documentation, we notify the client so their coding team can address them. Additionally, for coding mismatches where the model is not performing as expected, Maverick will take the necessary steps to improve the model and implement any needed customization ensuring the model is aligned with your organization’s specific coding policies before launch.

The Maverick autonomous coding solution offers a fully transparent platform designed to elevate performance oversight and strategic decision making. It features a comprehensive suite of detailed reports that enable monitoring of both model and coder accuracy and outcomes. Additionally, an interactive executive dashboard provides rea-time visibility into key performance indicators (KPIs), empowering leadership to make informed, data-driven business decisions.  

For example, the following metrics can be tracked independently by customers without vendor intervention:

• Automation Rate (Direct-to-Bill Rate): Customers can measure the percentage of claims automatically coded and submitted without manual intervention, giving insight into how effectively Maverick is streamlining the coding process.
• Accuracy: Accuracy can be tracked through internal audits and review processes, comparing coding outputs before and after implementation to measure improvements in compliance and correctness.
• Coder Productivity: Customers can monitor coder productivity by assessing the number of cases coded per hour or day, with Maverick customers typically experiencing a 50% increase in baseline productivity.
• Aging Days Within the System: Customers can track claim processing times and aging days to measure revenue cycle management efficiency, identifying any delays in reimbursement or issues related to denials.
• Variances between AI Model and Coder: Customers can use our very detailed layers comparison report for a code over code view into the code selection delivered by mCoder and, when applicable, the coder. This is important to the continuous fine tuning and improvement of the engine.
• Variances between Coder/Model and Auditor: Customers can use our very detailed layers comparison report to easily view audit/QA results to determine engine/coder accuracy and establish education feedback.

As a best practice we recommend that any payer specific billing rules are housed in the billing system and not in the Maverick platform. However, if you are unable to build and apply these rules in your billing system, Maverick can build payer rules in the coding platform to be applied to encounters. 

When new code sets are released, the Maverick team goes to work building any necessary crosswalks and updating the model with the information it needs to assign the new codes. This process ensures the engine is prepared to apply updated CPT®, HCPCS, and ICD-10 codes accurately and compliantly when they go into effect.

Data security is central to our solution. Maverick’s measures are designed to ensure our customers’ highest levels of data privacy and security. 

We maintain HIPAA compliance, and our platform is SOC 2 certified. All data is encrypted in transit and at rest to protect sensitive information from unauthorized access. We employ strict access controls and network segmentation to minimize security risks. Our cloud infrastructure, built and maintained in US located Amazon AWS servers, follows best practices for data security, including:

• Audit logging to track data access and modifications.
• Role-based access to ensure that individuals can only access data relevant to their roles.
• Continuous monitoring to detect and respond to potential security threats in real-time.

Success with autonomous coding requires more than just turning the system on. Organizations should focus on three key areas. By strengthening these foundational pillars before go-live, providers set the stage for higher direct-to-bill rates, stronger compliance, and long-term success.

• Workflow – Streamline processes from order entry through reporting to ensure accurate, complete inputs.
• Documentation – Standardize report templates and ensure required details are consistently captured.
• Coding Quality – Audit and correct coding practices, standardize policies, and provide ongoing education.

During the implementation process, we collaborate with your team to evaluate upstream processes such as order entry, report templates, and provider documentation. Cleaner and more standardized inputs make it easier for the model to assign correct codes the first time, minimizing exceptions and reducing the need for manual coder intervention.

Maverick customers typically go live within 90 days of contract signing.  The precise timeline depends on system integration, data readiness, and internal resource availability. We use a structured process which is led by an assigned project manager —discovery, integration, validation, training, and go-live—designed to achieve 85% DTB while maintaining 95%+ coding accuracy.

While the Maverick team leads the majority of the technical execution, your team contributes critical support by providing system access, historical coding data for validation, and input from key stakeholders. Maverick will integrate the platform with your system, according to agreed specifications. 

We conduct weekly implementation meetings and request participation from your project manager, IT director and your coding manager at a minimum. You’re welcome to include additional team members at your discretion.

After implementation, the Maverick team is highly involved in ensuring optimal performance. We continuously monitor direct-to-bill rates and coding accuracy to ensure sustained results. 

Performance and accuracy are safeguarded through a combination of auditing, monitoring, and client-driven quality checks:

• Go-Live Oversight
  At go-live, clients perform a 100% review of all cases coded by the engine, typically for one week. Based on those results, the client decides how many encounters can move to direct-to-bill (DTB) and what percentage they’d like to continue auditing as part of their internal QA. Those selected encounters are routed into a QA bucket for ongoing review.
• Close Monitoring in the First Weeks
  During the first week of go-live, Maverick’s Director of Coding Compliance monitors coder changes daily and provides feedback to the R&D team to make any necessary adjustments. After week one, monitoring continues on a weekly basis to ensure sustained accuracy and performance.
• Quarterly Audits
  Once the system is fully live, Maverick’s Director of Coding Compliance conducts quarterly audits for each client. This involves reviewing a random sample of 200 DTB encounters to confirm that accuracy is being consistently maintained over time.

This layered approach—validation, real-time monitoring, client QA, and quarterly audits—ensures that performance and accuracy are not only achieved at Go-Live but continue to strengthen over time.

Yes. While historical coding quality is critical, there are strategies to minimize the impact of poor data. As part of implementation, we carefully evaluate the historical data set during the validation process. If we detect errors or inconsistencies, we can flag those issues, exclude poor-quality portions of the data, and adjust the training set accordingly. We also benchmark the model’s performance against expert human coders to ensure it learns the correct coding patterns. This proactive approach allows us to minimize the impact of flawed data and still deliver a reliable, accurate model at go-live.