SonderMind, a national behavioral health organization with a network of more than 13,000 providers across all 50 states, has increased its capacity to process referrals while reducing administrative costs by automating intake and care coordination workflows.
The organization works with the Department of Veterans Affairs Community Care Network and large health systems, where referral accuracy and processing speed are critical. As referral volumes increased, SonderMind experienced growing backlogs, rising reliance on contractors, and error rates that required extensive manual review.
To address these pressures, SonderMind partnered with Magical to implement their agentic AI platform to automate VA referral intake. Since implementation, processing of monthly referral volume tripled in one region alone. During the same period, contractor spend was reduced by roughly 50%, while error rates on referrals fully processed by AI agents fell to under 1%, compared with an estimated 15% previously.
“Magical changed our ability to handle volume,” said Zach Holmes, VA account manager at SonderMind.
Before adopting agentic AI, SonderMind relied on a combination of manual workflows, temporary staff, and a legacy automation platform to move referrals from VA systems into internal workflows. That approach became less effective as volumes increased. Referral data often required full manual verification due to frequent mismatches in modality, state routing, and insurance information.
Delays were common during periods of high volume. While SonderMind previously converted most referrals within 24 hours, processing times stretched to several days as backlogs grew and staffing fluctuated. The gap between a referral arriving in the VA system and becoming a usable internal lead often went unmeasured but could take two to three days, slowing outreach and scheduling.
To keep pace with demand, SonderMind increasingly relied on temporary staffing, driving up administrative costs.
“With the amount of volume we were seeing, contractors became the default solution,” said Bianca Estrella, director of care coordination. “It was expensive and difficult to sustain.”
Under the new model, autonomous AI agents handle the full VA referral workflow, including identifying new referrals in VA systems, extracting structured and unstructured data, creating records in Salesforce, and routing referrals into downstream scheduling processes. The end-to-end workflows include verification steps that flag incomplete or ambiguous information for human review rather than forcing potentially incorrect entries.
“Manually, we were processing about 11 referrals an hour,” Estrella said. “During backlog periods, that increased to around 35 an hour with the automated workflow.”
SonderMind leaders cited reliability and error prevention as key considerations in adopting the technology. Rather than replacing human oversight, the system escalates uncertain cases to staff, reducing the likelihood of mismatches that can result in denied or unbillable care.
“I was concerned about AI introducing the same types of errors we see with manual work,” Holmes said. “Instead, it flags what it’s unsure about, and we don’t see the same kinds of mismatches anymore.”
Implementation took place over several weeks. Live referrals began processing shortly after onboarding, and the VA intake workflow stabilized within the first month as teams aligned on definitions of completion and exception handling.
“Most vendors want you to adapt to their product,” Estrella said. “This was built around how our referrals actually move through the system.”
Following implementation, SonderMind reduced contractor headcount significantly, lowering monthly spend substantially while increasing referral capacity. Processing times for clean referrals were reduced from multiple days to under an hour, enabling faster outreach and scheduling for veterans.
“What used to take days now happens in a much shorter window,” said Sean Simler, military operations lead. “That has a direct impact on how quickly veterans get connected to care.”
SonderMind is now extending the automated workflows to additional VA-related processes, including re-authorizations and status updates, and is preparing to support higher referral volumes from health system partners.
To learn more about Magical’s agentic AI platform and how health systems are scaling clinical and revenue cycle operations, visit www.getmagical.com.
