# Fern Cowan launches DeepCura's ledger-backed AI workforce for medical practices > Source: > Published: 2026-08-24 17:21:12+00:00 # Fern Cowan launches DeepCura's ledger-backed AI workforce for medical practices **The bootstrapped healthcare AI vendor is expanding from scribing into reception, scheduling and chart updates, with clinicians approving consequential actions.** By [RuntimeWire Staff](/author/runtimewire-staff) ยท Published Primary source: [PR Newswire](https://www.prnewswire.com/news-releases/deepcura-unveils-the-electronic-health-workforce-ehw--the-step-beyond-the-ehr-302858389.html) ## Why it matters Cowan is betting small practices will buy one supervised AI workforce instead of separate scribe, phone, scheduling and population-health tools. The ledger is his answer to AI's accountability problem. [Fern Cowan](https://deepgram.com/podcast/aiminds-006-fern-cowan-of-deepcura?ref=runtimewire) unveiled [DeepCura's Electronic Health Workforce](https://www.prnewswire.com/news-releases/deepcura-unveils-the-electronic-health-workforce-ehw--the-step-beyond-the-ehr-302858389.html?ref=runtimewire) on August 24th, expanding the healthcare AI product from an ambient documentation tool into a supervised group of digital receptionists, scribes, intake nurses, schedulers and inbox workers. The Electronic Health Workforce, or EHW, is available through [DeepCura](https://www.deepcura.com/?ref=runtimewire) from $129 per provider each month. DeepCura says it can connect with existing electronic health record systems, including Epic, athenahealth, eClinicalWorks and AdvancedMD, or operate as the main working system for a medical practice. Clinical changes require a clinician's approval, while administrative tasks follow a preview-confirm-execute process. Cowan came to healthcare software by an indirect route. In a 2024 Deepgram interview, he described moving from culinary arts into videography and advertising before teaching himself to build software with help from ChatGPT and contractors hired through Upwork. He chose healthcare partly because he saw it as a defensive industry while the advertising market weakened. That pragmatic origin still shapes his pitch. Cowan is building for small practices that lack the budgets and administrative departments of hospital systems, while using DeepCura itself as a demonstration of how much work a small human staff can hand to AI. DeepCura's [About page](https://www.deepcura.com/resources/about?ref=runtimewire) says three people and seven AI agents operate a product used by more than 6,000 clinicians. Those adoption and staffing figures are DeepCura's own. So are its claims of profitability, more than 1 million clinical encounters and more than two hours saved per clinician each day. Still, the design behind the EHW is a substantive expansion beyond the ambient scribe category where DeepCura started. ### A medical chart built like a ledger Cowan's central product decision was to borrow the append-only ledger from finance. Each call, appointment, payment, chart update and AI action becomes an entry on a unified timeline. Corrections are recorded as new events instead of overwriting the old information. DeepCura also borrows the idea of a commit from software version control. After a visit, the clinician can review the specific changes proposed for the chart rather than reading the full record again. AI-generated facts retain their provenance, including the conversation, recording timestamp or document page that produced them. Cowan calls this a "system of evidenced action." The phrase is marketing, but it describes a concrete attempt to solve a growing healthcare AI problem: once software begins performing multi-step work, medical practices need to identify the actor, the action, the source material and the human who approved it. "The EHR was built to remember. The EHW is built to work," Cowan said in the announcement. DeepCura says each AI worker has a name, defined privileges and a permanent activity record. Proposed clinical facts remain unverified until a clinician approves them. The EHW is designed to keep the source attached to a fact for the life of the record, down to the relevant words in a recording or the highlighted section of a fax. That architecture matters because approval alone does not settle accountability. A clinician's click can establish who accepted an AI-generated change, but legal and clinical responsibility still depends on the surrounding workflow, the quality of the evidence and whether the clinician had a realistic opportunity to review it. A [2026 paper on agentic AI governance in healthcare](https://arxiv.org/abs/2601.15630?ref=runtimewire) argued that organizations deploying multiple agents need identity registries, runtime policy controls, bounded access to health information, credential revocation and audit logging. DeepCura's ledger, scoped privileges and permanent records address several of those controls at the product level. Medical practices will still have to determine how those controls operate in day-to-day care. ### From writing notes to running the practice DeepCura's broader bet is that ambient documentation becomes a gateway into the rest of the medical office. An AI scribe already hears the visit. Cowan wants the same system to update structured records, identify care gaps, prepare outreach, manage incoming messages and schedule the next appointment. DeepCura launched its Ambient Data feature in July 2026. The feature converts values spoken during a visit into coded, trendable chart entries after clinician approval. A physician mentioning an A1c result, pain score or specialty-specific measurement can create a longitudinal record without entering the data separately. Those entries feed DeepCura's Population Panels, which let practices group patients by condition, recent measurement or care gap. The EHW also brings faxes, emails, calls and text messages into an AI-triaged inbox. A planned feature called Loops is intended to track open obligations, such as an expected test result or an unanswered referral, until the task is closed. DeepCura is trying to collapse software categories that medical practices often buy separately: scribing, intake, phone reception, scheduling, communications, structured charting and population health. The $129 entry price is part of that argument, although AI work beyond the subscription is metered through usage credits. The competitive market is moving in the same direction. [Parallel raised $20 million in March](https://www.beparallel.com/news/parallel-raises-20m-to-accelerate-ai-agent-deployment-in-hospitals?ref=runtimewire) for agents that automate hospital coding, billing and admissions work. In August, revenue-cycle vendor [R1 agreed to acquire Humata Health](https://www.r1rcm.com/newsroom/r1-to-acquire-humata-health-enhancing-phare-os-with-ai-powered-prior-authorization-automation-and-payer-provider-collaboration?ref=runtimewire), adding AI-based prior authorization to its operating system for healthcare providers. DeepCura is approaching that shift from the smaller end of the market. Cowan said in the 2024 interview that individual providers and small practices were easier for a young vendor to serve than hospitals with long procurement and compliance reviews. The EHW preserves that route while giving DeepCura a larger product to sell once customers begin with documentation. ### Cowan's bootstrapping thesis gets a product test Cowan has argued that small teams can outbuild larger organizations when agents perform implementation, support and operational work. DeepCura says its own agents answer sales calls, configure customer workspaces and build AI receptionists for medical practices. That internal use is evidence of Cowan's conviction, rather than independent proof that the model will work for every clinic. Medical offices carry different staffing rules, patient populations, state requirements and risk tolerances. A workflow Cowan is comfortable automating inside DeepCura may require tighter supervision in a practice handling patient communication or clinical data. DeepCura's funding story also contains a discrepancy. Its current materials describe DeepCura as profitable, fully bootstrapped and never funded by outside investors. A [2024 summary of a separate Cowan interview](https://rewskidotcom.substack.com/p/fern-cowen-founder-and-chief-architect?ref=runtimewire) said DeepCura's early evolution included a partnership with an angel investor. That older account conflicts with DeepCura's current assertion that it has never taken outside capital. Bootstrapping is integral to Cowan's product argument. DeepCura says lower headcount lets it sell a broader package at a lower price than heavily funded rivals. The EHW now has to prove that the same lean structure can support a larger operational role without weakening implementation, reliability or clinical oversight. Cowan has spent three years moving DeepCura from note generation toward practice automation. The EHW packages that progression into a single thesis: every action performed by AI should carry a source, a cost, a permission and a human decision where the stakes require one. That is a more credible direction than asking medical practices to trust an invisible agent. The test begins when clinicians have to review those receipts during a busy workday.