“Concurrence has been an awesome AI partner to us, especially in Care Product and Quality, designing, building, and deploying agentic solutions to increase our patient outreach capacity and make our coaches' workflows more efficient.”

Jay Srivastava
Chief of Staff, Diverge Health
The Company
Most of what keeps a Medicaid member from getting care happens nowhere near a doctor's office. As a value-based care enabler serving CMS populations, Diverge partners with health plans and independent primary care practices to help members manage chronic conditions, coordinate care, and access essential social services.
At the heart of that model are Diverge's health coaches across Michigan, Ohio, New York, and Illinois, working directly with members to turn their needs into practical next steps. That can mean finding an in-network specialist who is actually taking new patients, confirming what a member's dental or durable medical equipment benefit covers, arranging the non-emergency ride to an appointment, or pointing someone to a food pantry near their home.
To support that work, Diverge partnered with Concurrence to build an AI Care Navigation Assistant, starting with Medicaid members in Michigan. Shaped by Diverge's clinical and care management expertise, the agent brings health plan benefits, live provider search, and community resources into one conversational workflow. The goal is to put relevant information at coaches' fingertips so they can focus on the member and move toward an actionable next step without starting every search from scratch.

A value-based care enabler partnering with health plans and independent primary care practices serving CMS populations.
The Challenge
For the Medicaid members Diverge serves, care navigation extends beyond finding a provider. Transportation barriers, food insecurity, and housing instability can all affect whether a member gets to care or follows a treatment plan. Diverge's coaches address these Social Determinants of Health (SDOH) alongside clinical care, a whole-person approach that requires navigating information across health plans, provider networks, and community services.
A member's non-emergency transportation benefit might sit in a lengthy plan handbook while their dental coverage appears in a separate guide. Finding a nearby food pantry meant searching community directories, and locating an in-network specialist accepting patients meant working through external payer portals. Even with a plan-specific spreadsheet in the patient profile, coaches still had to consult the underlying resources and translate what they found into an actionable next step.
That work meant moving among per-plan PDFs, payer websites, benefit spreadsheets, and external directories, with nothing carrying over from one search to the next. Experienced coaches reported 15 to 20 minutes to locate a single resource. A member with several needs could require 30 to 45 minutes of research, and complex requests such as durable medical equipment could take up to two hours.
Food and transportation were the recurring needs. In areas such as housing, limited service availability added complexity that faster retrieval alone could not resolve. Diverge saw an opportunity to give coaches a more connected way to navigate all of it, supporting their expertise and creating more room for direct member engagement.
“Concurrence is a great AI development partner. Their team is creative, fast, and nimble. They helped us develop a care navigation agent that helped navigate benefits and provider searches in record time, saving our team time and quickly solving a problem that would have taken months to address via a traditional non-AI product development process.”

Meghan Sherlock Salazar
Senior Vice President, Population Health, Diverge Health
The Solution
Diverge and Concurrence built the agent to work alongside health coaches rather than communicate directly with members. The assistant supports three connected navigation workflows, with Diverge's care teams shaping how information is presented and used.
1. Plan Benefits Navigation
The initial work focused on benefits for Meridian and UnitedHealthcare Community Plan members in Michigan. Coaches query the agent in natural language, and it is designed to return grounded answers, translating everyday phrasing such as “skin doctor” into the corresponding clinical category, dermatology.
Phone numbers, addresses, and benefit details are drawn from source material, with document references and page numbers where available. Answers stay within the relevant plan context. When information cannot be located, the agent acknowledges the limitation and directs the coach to Member Services.
The agent's scope excludes diagnosis, triage, and clinical advice, with emergency and self-harm language directed toward 911 and 988.
“As a physician, my first question about any AI touching our members was where the guardrails were. Concurrence took that seriously from the start, built the agents to stay inside clinical and regulatory lines, and raised the compliance questions before we had to ask them. The result is something I'm comfortable putting in front of our coaches and our members, which is not a small thing in our member population of Medicare and Medicaid recipients.”

Vincent Pantone, MD
Chief Medical Officer, Diverge Health
2. Live Provider Search
Diverge extended the agent beyond static plan documents into live directory searches covering 30+ care types across payer directories, layered with Diverge's vetted primary care provider list, so coaches can connect plan coverage with real options in the member's community.
To reduce the risk of outdated referrals, the implementation combines Meridian's live public provider finder with a March 2026 directory fallback of 42,000+ providers, while UnitedHealthcare searches use live Find Care guest flows. In rural counties, ZIP code searches could be too broad, but sending member street addresses to commercial mapping APIs raised privacy concerns. Concurrence addressed this with a private Nominatim geocoding service and a dedicated Michigan index from OpenStreetMap, keeping address processing inside its controlled cloud environment.
Early usage bears out the investment. Provider and specialist search is the workflow coaches reach for most, named by 60% of surveyed coaches as their most recent task.
3. Community Resource Navigation
The agent draws on FindHelp for community resources across food, housing, goods, transit, and legal aid. Bringing these searches alongside benefits and provider information supports the whole-person approach at the center of Diverge's care model. A coach can explore the different needs a member raises through a single entry point, then use their expertise to guide the conversation and determine the next step.
Co-Development and Pre-Launch Validation
Diverge's clinical and care management team helped shape the agent's development and pre-launch validation. Six Diverge team members evaluated 150+ real-world scenarios across 24 benefit topics and both health plans, bringing frontline expertise directly into the process.
Testing covered everyday coach shorthand, questions spanning multiple benefits, and safety boundaries, including out-of-scope questions, prompt injection, and requests for clinical advice. Feedback on provider details, next steps, and resource retrieval informed later updates, including work to expand provider search and pharmacy coverage. That collaboration kept the work centered on practical care-team needs: relevant information, clear next steps, and an assistant with a defined role.
“Their collaborative approach, commitment to real-world validation, and thoughtful attention to member experience, consent, and data privacy have been instrumental in building solutions that meet our operational needs while keeping members at the center.”

Jill Howard
Senior Director, Care Product, Diverge Health
The Outcomes
The agent brings previously fragmented navigation tasks into one assistant across two payer networks and an expanded provider-search scope.
More Time for Direct Member Support
- 52% reduction in time per member inquiry. Coaches reported 19.6 minutes of manual research before the agent. With it, they reported 5.4 minutes to get an answer plus 3.9 minutes checking it, a net 9.3 minutes per need.
- An average of 10 minutes in coach time returned per inquiry, redirected to direct member engagement.
- Verification time is counted against the agent, not excluded. The 3.9 minutes coaches spend checking an answer is held inside the 52%. Looking only at the time to get an answer, the reduction is larger.
Coach Adoption and Scale
Finding providers and specialists was the most common use, named by 60% of surveyed coaches as their most recent task, followed by transportation at 20%, then housing, urgent care and nurse line, and behavioral health and care management.
What's Next
The Michigan pilot was deliberately small. The next phase is about volume, reach, and who gets to ask the questions.
- Ohio. Diverge and Concurrence are extending the agent to health coaches in Ohio, with plan information and community resources tailored to the new market. A larger coach base and a second set of payer networks should move it from a pilot-scale tool to part of the daily workflow, which also gives both teams the lookup volume needed to measure capacity returned, not just time saved per need. New York and Illinois follow.
- Return should grow as trust does. Coaches currently spend about 4 minutes verifying each answer, which is the right instinct early on. As the agent's track record builds across more lookups and more plans, we expect the amount of checking to shrink for the question types coaches have seen the agent get right repeatedly, which raises the time returned per need without any change to the agent itself.
- Medicaid and Medicare together. Diverge serves both populations, and much of what makes navigation hard (fragmented benefit documents, live network churn, thin local service availability) looks the same on the Medicare side. Extending the agent across both lines of business would let one assistant cover a coach's whole panel instead of part of it.
- A patient-facing path. Today, the agent supports coaches. Putting a version in front of members could let them self-serve the questions that do not need a coach, and transportation is the clearest case. A member who needs a ride to an appointment wants the benefit details and the booking number now, not at the next check-in. Self-serve on the routine questions frees coaches for the work that genuinely needs a person.
In parallel, the partnership is advancing a separate outreach and re-engagement program, moving from an initial Medicare voice pilot toward a text-forward, multi-touch model for Medicaid members.
“Concurrence brought real depth in AI agent builds and a team that met us where we were. They built our Medicaid engagement scripts instead of forcing us into a premade template. They also walked us through the regulatory and compliance landscape for texting before we wrote a single message. They have been true partners from start to finish.”

Sara Frech
Senior Director, Quality & Risk Adjustment, Diverge Health
These efforts share a purpose of making it easier for care teams to connect members with the support they need. By bringing its frontline expertise into the design of AI-supported workflows, Diverge is pursuing more than administrative efficiency and creating more room for the human side of care.