A $45 Million Bet on Empathy at Scale
Ellipsis Health closed a $45 million Series A in June 2025, as PR Newswire reported, led by Salesforce, Khosla Ventures, and CVS Health Ventures — a platform vendor, a thesis‑driven venture firm, and a payer‑pharmacy giant backing a voice‑AI care manager called Sage. Mitsui Global Investment, Collier, E12, and AME Cloud Ventures also participated. The technology that built its reputation detecting depression from 60 seconds of speech is moving from screening into longitudinal care management.
Sage runs on Ellipsis's proprietary Empathy Engine, trained on millions of live clinical patient calls and built atop patented vocal‑biomarker technology that analyzes both what patients say and how they say it — acoustic patterns, tone, timing. The system adjusts its own tone and approach in real time based on a patient's emotional state, a design choice meant to sustain engagement with the high‑risk, high‑cost members who drive the bulk of healthcare spending. Co‑founders Mainul Mondal and Michael Aratow have argued since the company's founding that mental health deserves the same clinical priority as physical health; Sage extends that premise into longitudinal care management.
The product integrates with Salesforce Health Cloud's workflow engine. That integration lets health plans and providers embed Sage into existing case‑management operations without replacing their CRM. Ellipsis says the integration alleviates staffing shortages and rising costs by expanding capacity for proactive patient outreach — triage, care coordination, disease management, the between‑visit phone work that has always been essential but has become harder to scale as clinician burnout and patient complexity climb.
The Distribution Engine
The Health Cloud integration is the distribution lever that turns Sage from a promising pilot into an enterprise‑grade option for health plans and providers already running on Health Cloud. When Amit Khanna, SVP and GM of Health & Life Sciences at Salesforce, said the future of healthcare requires AI that extends human capabilities while preserving empathy, he was describing the integration path Sage now travels — native to the workflow engine that care managers use daily.
CVS Health Ventures did not write a check for strategic signaling alone. The payer operates a care‑management organization staffed by more than 12,000 nurses. Nathan Frank, Chief Digital and Technology Officer at Aetna (CVS Health's insurance business), described how those clinicians sit in pods with UX designers and software engineers to remove friction from the very between‑visit conversations Sage automates. Frank noted that ambient listening during calls already summarizes information for future visits, and that one AI tool alone returned 90 minutes a day to clinicians. Sage enters that same environment as an agent that links directly into the Health Cloud workflow engine, pulling patient context, writing back encounter notes, and triggering follow‑up tasks without forcing a nurse to switch screens.
Frank emphasized that CVS does not use AI to deny care — only for auto‑approvals, administrative burden, and experience. He also warned that frontier AI companies often lack the healthcare experience, data, and interoperability to make their products work in production. The Health Cloud integration sidesteps that gap: Sage arrives pre‑integrated with the Health Cloud data model, the same security controls, and the same change‑management playbook that Salesforce customers already follow.
The CMS Health Tech Ecosystem partnership, announced in July 2025, adds a federal distribution channel. Ellipsis was selected as a key partner in the agency's digital modernization program, pledging to develop conversational AI that integrates with health records for personalized patient support with full privacy protection, plus specialized tools for diabetes and obesity management that know when to escalate to a human professional.
Earlier partnerships illustrate the pattern. Tiatros, a digital health platform serving employers, health systems, payers, and government agencies, integrated Ellipsis vocal‑biomarker technology in 2022 to triage patients, check care needs between sessions, inform treatment plans, and monitor results. Ceras Health followed the same year, embedding real‑time severity scores for anxiety and depression into its clinical monitoring platform to improve triage for the more than 18 million Medicare patients, per PR Newswire's data, with mental‑health conditions. Both deals proved the technology could sit inside another vendor's platform and surface actionable data at the point of care.
Hal Paz, a former health system CEO now at Khosla Ventures, put the operational problem bluntly: voice‑based care management has become essential but is nearly impossible to scale cost‑effectively without significant loss of quality.
Sage answers that by handling the time‑intensive outreach, assessment, and coordination tasks that currently consume nursing hours, then handing off only the complex decisions that require clinical judgment.
Board data shows Ellipsis hiring for the scale that partnership demands: VP Payer Sales and VP Provider Sales roles listed at $160k–$220k base, plus senior ML and backend engineers building the agentic AI evaluation and forward‑deployed infrastructure that keeps Sage reliable inside enterprise environments. Those hires map directly to the sales motion the Health Cloud integration unlocks and the clinical integration work the CMS and CVS partnerships require.
The Scribing Giants and the Care‑Management Gap
Microsoft's 2021 acquisition of Nuance for roughly $16 billion set the baseline for today's voice‑AI arms race in healthcare. The combined entity now claims Dragon Medical One runs in 77 percent of U.S. hospitals and serves more than half a million physicians worldwide. By March 2025 Microsoft said DAX Copilot was logging over 3 million patient visits a month across 600 health‑care organizations. Northwestern Medicine reported that physicians using the tool in at least half their encounters added 11.3 patients per month on average, cut note‑taking time 24 percent, and reduced after‑hours "pajama time" 17 percent. Dragon Copilot, the rebranded successor, began general availability in the United States and Canada in May 2025 with a European rollout slated for the months after.
| Company | Primary focus | Known funding | Key product |
|---|---|---|---|
| Microsoft / Nuance | Ambient scribing, clinical documentation | $16B acquisition (2021) | Dragon Copilot, DAX Copilot |
| Abridge | Ambient scribing, structured notes | $460M+ | Abridge Inside |
| Suki | Voice assistant, clinical documentation | ~$170M | Suki Assistant |
| Ellipsis Health | Care management, patient engagement | $45M Series A (2025) | Sage |
Abridge and Suki have raised $460 million and $170 million respectively. Both target the same ambient‑scribing workflow that DAX Copilot dominates. Sage enters a different lane: care management rather than documentation. Ellipsis positions Sage as an "emotionally intelligent" care manager that adjusts tone and approach based on a patient's mental‑health signals — something the scribing tools do not attempt.
The company's CMS Health Tech Ecosystem selection signals federal comfort with that distinction.
Market pressure is measurable. Google Cloud found clinicians spend nearly 28 hours a week on administrative tasks. The KLAS Arch Collaborative identified academic‑center physicians as the second‑most burned‑out cohort, citing complex workflows that drive after‑hours charting. In 2023 roughly 53 percent of physicians reported burnout. Ellipsis's own Series A materials argue "it has become essential, but it's a challenge to scale cost effectively or without significant loss of quality" — a direct shot at the manual call‑center model still used by most health plans.
Hiring signals on Zero G Talent's board reinforce the competitive posture. Nuance Labs posted a single role in the past seven days, a Research Assistant, with a board‑wide salary band of $42,000–$62,000. Ellipsis Health, by contrast, listed 14 salaried openings in the same window spanning ML engineering, backend infrastructure, forward‑deployed engineering, and two VP‑level sales roles (payer and provider) with bands of $160,000–$250,000. The disparity suggests Ellipsis is staffing for go‑to‑market expansion while Nuance, now inside Microsoft, is recruiting narrowly for research.
Proxy indicators are consistent: CMS partnership traction, payer‑backed Series A capital, and a documented clinician‑time crisis that scribing alone does not solve. Microsoft's Dragon Copilot rollout across five countries by late 2025 will test whether a documentation‑first platform can stretch into longitudinal patient engagement. Sage bets it cannot — that the workflow, data model, and regulatory posture for care management are distinct enough to sustain a separate category leader.
The Math at the Bedside
Ellipsis Health's Sage enters that equation with a specific claim: expand staffing capacity immediately and reduce operating costs through proactive, high‑quality patient engagement. The mechanics rest on integrating with the Health Cloud workflow engine so care managers can offload routine outreach without leaving their existing console. That integration is the operational hinge; without it, another point solution creates more friction than it resolves.
For providers, the staffing arithmetic is brutal. The National Alliance on Mental Health estimates, according to PR Newswire, 55 percent of people with mental illness receive no treatment, and the U.S. faced a projected shortage of up to 30,000 psychiatrists by 2024. Primary care physicians already managing complex chronic patients, roughly one‑third of whom also battle depression, cannot absorb behavioral‑health screening on top of 15‑minute visits. Sage's pitch is that its Empathy Engine, trained on clinical calls, conducts the initial engagement, triages acuity, and surfaces the patients who need human intervention now. "Practice at the top of their license" is the industry phrasing; in practice it means a care manager handles more patients because the AI handles the check‑ins that don't require clinical judgment.
Patient engagement metrics remain the black box. The company says Sage adjusts its approach based on emotional state detected from vocal biomarkers, acoustics and semantics from under 60 seconds of speech, and that this "meets patients where they are."
Early partners tell a directional story: Tiatros, whose platform serves employers, health systems, and payers, described the combination as a "force multiplier that enables systems of care to dramatically increase the supply of mental health services, shorten wait times, reduce the overall cost of care and overcome health disparities." Ceras Health, focused on Medicare transitions of care, blends Ellipsis voice analysis with 24/7 clinical teams and remote monitoring to personalize outreach for vulnerable populations. This selection signals federal interest but the program's own milestones are still forming.
The engineering side of the operation is visible in hiring. Ellipsis Health's board listings show 14 salaried roles with a median band of $200,000: Senior ML Engineers in Agentic AI and AI Evaluations at $175,000–$250,000, Senior Backend Engineers across infrastructure and forward‑deployed roles at the same band, plus VP Payer Sales at $180,000–$220,000 and VP Provider Sales at $160,000–$210,000. That mix, heavy on evaluation infrastructure and forward‑deployed engineers who sit inside customer environments, matches a product moving from model development to production hardening at enterprise scale. By contrast, Nuance Labs, a legacy voice‑AI incumbent now inside Microsoft, lists a single recent role at a $42,000–$62,000 band, suggesting a different hiring posture. The salary differential reflects where the market sees the hard problems: not in speech recognition, but in longitudinal care logic, safety guardrails for empathetic AI, and the integration plumbing that makes a voice agent a billable part of a care plan.
The economic case for payers hinges on two gaps Ellipsis has quantified since its 2021 Series A: the screening gap (identifying who needs help sooner) and the monitoring gap (visibility between appointments). Cigna's StressWaves Test proved the acoustic‑semantic model could stratify stress at scale. Sage extends that into longitudinal care management — the difference between a snapshot and a film strip. If the CMS partnership yields measurable reductions in ER utilization or readmission for dual‑eligible populations, the cost‑avoidance math becomes concrete. Until then, the operational outcome is a deployed system inside Salesforce Health Cloud with live patient conversations happening today, and a hiring plan built to make the next 10x scale work without the quality loss that has plagued previous attempts to automate care management.
Where the Money Meets the Rules
The FDA's definition of digital biomarkers — "a characteristic or set of characteristics collected through digital health technologies, which serve as indicators of normal biological processes, pathogenic processes, or responses to exposure or interventions", places Ellipsis Health's voice vital sign squarely in regulatory territory that remains unsettled. Speech features recorded through microphones qualify as digital biomarkers under that framework, but the agency has not yet established a dedicated clearance pathway for voice‑based mental‑health diagnostics. Ellipsis's Series A announcement in 2021 positioned its technology as pioneering "a new clinical standard in mental health care," yet the regulatory strategy since then has been incremental: clinical validation studies, payer pilots, and integration with existing care‑management workflows rather than a direct FDA submission for the voice biomarker itself.
That caution reflects the evidence gap. A 2025 Nature Digital Medicine review of digital‑biomarker studies found that only two incorporated external validation and three performed model calibration, numbers that would give any FDA reviewer pause. This partnership signals a different regulatory vector: rather than seeking a standalone device clearance, Ellipsis is co‑developing such AI under CMS oversight, with such tools.
This approach treats the voice biomarker as a component of a broader care‑management system, potentially a SaMD (Software as a Medical Device) accessory, rather than a diagnostic device requiring de novo or 510(k) clearance.
The hiring plan tells the same story. Zero G Talent's board shows 14 open salaried roles at Ellipsis as of the latest ingest, clustered in San Francisco hybrid positions: Senior ML Engineer (Agentic AI), Senior Backend Engineers across evaluations, infrastructure, and forward‑deployed functions, plus VP‑level sales leads for payer and provider channels. Salary bands run $175,000–$250,000 for engineering and $160,000–$220,000 for sales leadership. The engineering roles, particularly "Agentic AI" and "AI Evaluations", point to a product roadmap focused on autonomous care navigation and rigorous model monitoring, not regulatory submission packaging. The forward‑deployed backend role suggests customer‑facing implementation capacity, consistent with the Health Cloud distribution model and CMS pilot expansion.
Product milestones over the next 12–18 months will likely center on three vectors. First, deepening the Salesforce Health Cloud integration: Sage already links with the workflow engine; the next layer is bidirectional data exchange with EHRs so that voice‑session insights trigger care‑plan updates in real time. Second, the CMS‑pledged diabetes and obesity modules, conversational agents that escalate to human clinicians based on voice‑detected distress or adherence signals. Third, multilingual and acoustic‑robustness work: the current voice vital sign was validated primarily on English‑speaking U.S. populations; scaling to Medicaid and Medicare Advantage populations demands broader language support and noise‑resilient models.
The longer‑term outlook hinges on reimbursement. The projected 30,000 psychiatrist shortage creates pressure for novel reimbursement mechanisms. Ellipsis's strategic investors, CVS Ventures (payer‑pharmacy integration), Salesforce (workflow platform), Khosla Ventures (deep‑tech patience), each hold leverage points: CVS through Aetna MA plans, Salesforce through Health Cloud's payer install base, Khosla through FDA advisory experience from portfolio companies. If any of those channels secures a coverage decision for voice‑enabled care management, the regulatory conversation shifts from "is the biomarker valid?" to "is the care model reimbursable?", a fundamentally different, and more tractable, question.
Sage already conducts patient interactions. The technology that once lived in a research lab now sits inside the workflow where decisions get made. The next conversations will tell whether empathy can scale.
Working in frontier tech? Zero G Talent tracks the openings: see every open Nuance Labs role, browse frontier tech jobs, openings at Ellipsis Health, and the people building the field.