mdhub pairs AI workers with an EHR built for behavioral health
There is a version of the mental health crisis that rarely makes the news, and it happens at 9pm on a weeknight. A therapist who finished her last session two hours ago is still at her desk, typing up progress notes, chasing an insurance verification, and answering intake messages from people who called while she was in session. She went into the field to treat patients. A large share of her working life is spent feeding software.
The founders of mdhub put a number on it when they launched: clinicians in behavioral health lose roughly half their time to administrative work. Whatever the exact figure at any given clinic, the direction is not in dispute. Demand for mental health care keeps climbing, the supply of clinicians is not keeping up, and a meaningful fraction of the capacity that does exist gets burned on documentation, scheduling, intake, and billing.
mdhub's answer is more ambitious than a dictation tool. The company, founded in 2023 by Dominik Middelmann and Efrén A. Lamolda and part of Y Combinator's Summer 2024 batch, describes itself as an AI-native operating system for behavioral health clinics. In practice that means one platform that combines AI workers with a purpose-built EHR, CRM, and revenue cycle management, so the clinical team and the AI operate in the same system instead of passing files between a half-dozen disconnected tools.
The paperwork tax on care
Behavioral health has an unusual relationship with its software. Most clinics run on EHRs that were designed around general medicine and, more specifically, around billing. The note formats, the intake flows, and the scheduling logic were built for fifteen-minute primary care visits, then adapted, badly, to fifty-minute therapy sessions and ongoing psychiatric care.
The mismatch shows up everywhere. Therapy documentation has its own conventions, with DAP, BIRP, and SOAP note formats that payers expect to see done properly. Intake is heavier than in most specialties, because a new patient typically needs screening forms, consent documents, and insurance verification before anyone can say whether the first appointment is even viable. Treatment is longitudinal, which means care plans and progress tracking matter more than in a specialty built on one-off visits.
Layer a staffing shortage on top and the result is predictable. Phone calls go unreturned because the front desk is underwater. Prospective patients sit on waitlists or give up. Clinicians finish notes at night, which is one of the most commonly cited drivers of burnout in the field. That is what happens when a high-touch specialty runs on software built for a different job.
From assistant to operating system
mdhub did not start out trying to replace the EHR. Its first product, launched out of YC, was a clinical AI assistant for mental health clinics focused on the most acute pain point: documentation. The assistant, which the company calls Emma, listens to a session and produces a structured clinical note, along with patient summaries and treatment plans, in the formats behavioral health actually uses.
Then the scope grew, and the progression is visible in the company's own launch history: first a clinical AI assistant, then AI agents for mental health clinics, then, this spring, the full platform. In April 2026 mdhub launched what it now sells: an AI-native operating system where the EHR, the CRM, and billing live in one place, with AI workers staffed across the workflows.
The named workers are the most concrete way to understand the product. Emma, the clinical assistant, handles documentation and gives clinicians back the hours they used to spend typing. Sarah, the AI admissions coordinator, runs the front of the funnel: she answers inquiries, verifies insurance, sends intake forms, and books patients, and she does it around the clock, with reminders and follow-ups, so a clinic no longer loses patients to an unanswered phone. Around them sits the rest of the clinic's machinery: scheduling, telehealth with HIPAA-compliant video and an AI scribe embedded in every visit, e-prescribing, and billing.
For the clinic, the pitch is a single system where the AI is staff, not a plugin. For the clinician, the pitch is simpler. See the patient, and the note is drafted before you get up from the chair.
Why the EHR had to be rebuilt
The obvious question is why mdhub builds the EHR at all. AI scribes are a crowded category, and the safer-looking play is to sell documentation software that sits on top of whatever system a clinic already uses.
The company's bet is that the integration layer is exactly where the value leaks out. A scribe bolted onto a legacy EHR can draft a note, but a person still has to move that note into the chart, update the treatment plan, trigger the follow-up, and make sure the billing codes match. Every handoff between tools is a place where work falls back onto a human. If the AI worker and the system of record are the same product, the loop closes: the note lands in the chart on its own, the intake Sarah completed flows straight into the clinical record, and the billing side sees all of it without anyone re-keying data.
That architecture is what "AI-native" means here, and it is the part a legacy vendor cannot easily copy. An EHR that was designed for human data entry can add AI features at the edges. It cannot easily become a system where software does most of the data entry, because its entire design assumes the opposite.
mdhub is pragmatic about the installed base, though. In October 2026 the company became a vetted partner on the athenahealth Marketplace, bringing its AI admissions automation to clinics that run on athenaOne. That reads as a wedge strategy: meet large clinics where they are, let Sarah prove herself on intake, and let the full platform make its own case over time.
Intake is where clinics lose patients
Of all the workflows mdhub automates, admissions may be the one with the clearest economics.
A behavioral health clinic's growth problem is rarely demand. It is conversion. Someone finally decides to seek help, calls a clinic during their lunch break, and reaches voicemail. The clinic calls back the next day and misses them. The intake packet goes out by email and sits unopened. Insurance verification takes another few days. Some meaningful share of those people, often the ones who most needed the appointment, drift away at each step. The clinic paid to acquire a patient it never saw, and the patient is back where they started.
Sarah attacks that funnel directly, because software does not have office hours. An inquiry at 10pm gets answered at 10pm. Insurance gets verified before the first conversation with a human. Forms go out immediately and get chased with follow-ups. The patient shows up to the first appointment with the paperwork done, which is better for the clinic's revenue and, more to the point, better for a person who found the courage to ask for help and got an answer instead of a hold queue.
This is also the use case where mdhub's numbers are most striking. The company says clinicians on the platform saved more than 100,000 hours of admin work within six months of launch, and as of this spring it reported more than 10,000 clinicians on the platform. Those are mdhub's own figures rather than audited ones, but the direction they point matches what the product is built to do: take the work that does not require a clinical license off the people who have one.
Founders who build close to the workflow
mdhub is a San Francisco company with a European-flavored founding story. Middelmann, the CEO, studied management at London Business School and started his career at BCG. Lamolda, the technical co-founder, met him in 2018 when both were working at a high-growth startup, several years before they started the company together in 2023.
What stands out from the outside is the discipline of the sequence. The company earned its way into clinics with a documentation assistant, the single feature clinicians were desperate for, and used that position to learn the rest of the clinic's operations before trying to own them. Plenty of startups pitch an "operating system" on day one and discover that nobody hands their whole practice to an unproven vendor. mdhub built trust in the exam room first, then expanded to the front desk and the billing office. The athenahealth partnership suggests the same patience: sell alongside the incumbents today, compete with them on the merits tomorrow.
The vertical bet
Zoom out and mdhub is a clean test of a thesis that is playing out across healthcare software: in specialties where the workflow is distinctive enough, the AI and the system of record want to be the same product.
Horizontal AI scribes treat documentation as a general problem and sell to every specialty at once. mdhub is betting the opposite way, that behavioral health is different enough, in its note formats, its intake burden, its longitudinal care, and its payer quirks, that a purpose-built system wins. The prize for being right is much bigger than a per-seat scribe subscription. The system of record is where the retention is, where the billing flows through, and where every future AI worker gets deployed first.
Behavioral health is also a sympathetic place to run this experiment. The specialty's shortage is measured in human suffering, not just margin. Software cannot train more psychiatrists. What it can do is stop wasting the ones we have on paperwork, and make sure that when someone finally reaches out for help, something answers.
mdhub's wager is that this gets built as one system rather than a stack of point solutions. Ten thousand clinicians in, by the company's own count, the wager is holding up.