Savista and Minset hit up to 93% touchless medical coding with audited accuracy

Savista said Wednesday its coding-automation partnership with Minset has reached up to 93% fully touchless medical coding while sustaining 100% accuracy in the audits that validated the rollout.

Hospital coding backlogs and a shortage of coders slow the cash and wear out the people who are left. An agent that can finish most charts on its own, with a thinner human check still in place, is the bet: more speed, without pretending the person has left the job.

On Wednesday, 30 September 2026, Savista announced results from its coding-automation partnership with Minset. Savista calls itself a healthcare operations company with more than 35 years in revenue cycle management. Revenue cycle management, often shortened to RCM, is the work of turning a patient’s visit into a claim a payer will pay. Medical coding is the step that reads the chart and picks the codes for the diagnosis and the visit. The release says the two companies work with a person still in the loop: the software supplies speed and scale, and a human coder supplies accuracy and judgment. Business Wire carries the release. A Business Wire reprint stamps it September 30, 2026, at 9:00 a.m. Eastern. The reprint does not print a city. The figures below are Savista’s.

Savista says the partnership has reached up to 93 percent fully touchless automation of medical coding. Touchless, in that sentence, means a chart coded with no person interacting with it. Ninety-three percent is the ceiling Savista states. It is not a rate the release prints for every kind of chart. Charts outside that share still need a person. Savista says that person validates the chart. The person does not code the whole chart by hand. Those lines are Savista’s.

Savista says it first checked accuracy by auditing every coded chart, which it states as 100 percent of them. It says that accuracy held at 100 percent. After that held, Savista moved to an ongoing audit of a 10 percent sample, about one chart in ten. Savista calls the smaller audit a sign the earlier result lasted. The release does not name an outside firm that ran the audits. The 93 percent and the 100 percent accuracy are Savista’s account of its own checks.

The release splits the automation by the kind of chart. Radiology coding, it says, is 93 percent automated. Evaluation and management leveling, which the release shortens to E&M, is up to 80 percent automated. An E&M code is the level of an office or hospital visit: how complex the visit was, which changes what the bill says. Savista says autonomous coding is in use across multiple specialties. It does not name those specialties beyond radiology and E&M. It does not name a hospital.

Minset’s side is a multi-skilled agentic AI platform. Agentic, here, means the software can take the next step itself, including assigning the code, rather than only suggesting one for a person to apply. The release says Minset’s agent handles full, touchless coding for a majority of the charts. Savista’s credentialed coding professionals are the quality-assurance layer. Quality assurance, the QA in the headline, is the check that the code is fit to send. A credentialed coder holds a coding certification. Savista says those professionals average more than 7.5 years of experience. The about box names Minset’s coding product as mCoder, for autonomous medical coding, next to mDenials for denials and m360 for patient engagement. A denial is a payer’s refusal to pay a claim.

Jan Grimm, chief executive of Savista, said the partnership is meant to give clients what they have always needed: faster turnaround, greater scale, and consistent accuracy, without compromising compliance. He said Savista’s coding teams are using Minset’s AI to process more work faster, with quality assurance built in at every step. For clients, he said, that should mean fewer backlogs, stronger revenue, and confidence that every claim meets the highest standard. That quotation is his, in the release.

Trey McMillian, chief executive of Minset, said the next healthcare platforms will not simply digitize work. They will create capacity that did not exist before. He said the partnership shows how autonomous AI can multiply the reach of expert teams while keeping the accuracy, oversight, and trust that healthcare demands. He called it more than a better way to code, and a base for a new way to run healthcare operations. That quotation is his, in the release.

Savista says it plans to extend the program to more specialties and more client sites in the coming months. It does not name those sites. Savista will be at the AHIMA 2026 conference from October 4 through October 6 in San Antonio, at booth 1119. AHIMA is the American Health Information Management Association, the group for people who manage health records and coding. The booth line invites attendees to talk with Savista about the partnership. It is not a new result.

The release’s about box says Savista works with hospitals, health systems, and physician practices on the cancer registry and the revenue cycle, including eligibility, coding, denials, and accounts receivable. Accounts receivable is money billed and not yet collected. It says the workforce holds more than 20 accreditations, including trainers certified on Epic. Epic is a widely used hospital records system. It says Savista serves more than 800 clients across 49 states. Those are Savista’s company figures. The release does not say which of those clients are on this coding-automation program. Minset’s about box says it built what it calls the first multi-skilled agentic AI platform made for healthcare revenue-cycle work. It says the founders came from Google, Microsoft, Optum, and Salesforce, and that HealthX Ventures backs the company. Those lines are Minset’s.

On minset.ai, Minset names the agent Mia. The homepage calls Mia the pilot of the revenue cycle, not a copilot, and says Mia replaces repetitive work. The mCoder page says Mia reviews the visit, applies the codes, and adjusts when a payer’s rules change. A payer is the insurer or other party that pays the claim. The page says mCoder connects to the electronic health record, the hospital’s chart system, or to a pipeline the customer already runs. It lists fewer coding errors, faster claims, and less rework. The homepage also prints its own outcome lines: more than 300 organizations, 40 percent fewer billing errors, a 25 percent increase in revenue, a 50 percent cut in denials, and 60 days until the spend pays for itself, which the page calls ROI. The homepage does not tie those lines to Savista, or to the 93 percent figure. Wednesday’s touchless rate, the radiology and E&M splits, and the audit story are on the Business Wire release. Neither page prints a price.

The picture is Minset’s mCoder graphic. A blue shield sits on a dark field, with a teal strip along the left edge and laurel branches on either side. The words on the graphic read “state of the art on MDACE.” MDACE is a public test of automated medical coding. It checks whether a system can point from a code back to the clinical note. Inside the shield, a page lists three steps: Clinical Note, Evidence Span, and ICD-10 Code. An ICD-10 code is the label for a diagnosis on a claim. An evidence span is the stretch of the note offered as the reason for that code. A bar along the bottom reads “mCoder · Clinical Note → Evidence → ICD-10.” Minset published this graphic with its May 2026 note on that test. It is a product graphic. It is not a chart from a named hospital, and the words on it are not the dateline of Wednesday’s release.

In plain terms, Savista said on Wednesday that Minset’s coding agent now finishes up to 93 percent of charts with no person touching them. Radiology is the 93 percent line. Visit-level E&M coding is up to 80 percent. A Savista coder still checks quality. Savista says a full audit of every coded chart held at 100 percent accuracy, and that it now audits about one chart in ten. The release does not name a hospital, does not name an outside auditor, and does not describe a medical-device clearance.