Suki, the Redwood City, California-based company that built its business on listening to patient-clinician conversations and turning them into structured clinical notes, announced Monday it raised $90 million in a Series E funding round — and made clear the money is not going toward better note-taking. The company's next chapter, CEO and founder Punit Soni told reporters, is autonomous AI that keeps working after the physician leaves the exam room: scheduling the cardiology consult, submitting the prior authorization, notifying the patient — without any additional physician input.

That bet puts Suki at the edge of a genuinely new category in healthcare AI, and in territory where the regulatory and liability frameworks have not caught up with the technology.

The Round

Healthier Capital leads Series E, a healthcare-focused growth fund. Partner Aman Mahajan, MD, PhD, MBA — whose background includes operational roles at UPMC and UCLA Health — will join Suki's board of directors, making the investment explicitly strategic as well as financial. New investors Crosslink Capital and Hercules Capital also participated.

Hercules Capital's presence is worth noting specifically. The firm is a publicly traded business development company that combines debt and equity financing, and it frequently backs pre-IPO technology companies between late-stage venture funding and a public offering. Its Hercules Capital BDC lending model suggests Suki may be approaching the window in which debt financing complements equity — a conventional signal in pre-IPO positioning.

All of Suki's returning institutional investor list came back for the round: Venrock, Flare Capital Partners, Hedosophia, Philips Ventures, and Zoom Ventures. Philips Ventures, an arm of the clinical monitoring and imaging giant, gives Suki a potential pathway into hardware-integrated workflows well beyond the physician's desktop. Zoom Ventures' continued presence reflects a deeper tie: Suki has integrated its ambient note engine into Zoom Workplace for Clinicians, turning the video-conferencing platform into a direct distribution channel rather than simply a customer.

The Series E brings Suki's total capital raised to more than $255 million. The company did not disclose a post-money valuation.

Why a $90M Round Is a Strategic Statement

The number itself carries a message. Abridge, Suki's closest competitor in the ambient AI category, secured its $300M Series E valuation in 2025 at $5.3 billion. Microsoft's Nuance, which sells the Microsoft Nuance DAX Copilot ambient scribe, has the distribution weight of Microsoft 365 and deep health-system relationships built over two decades.

Against that backdrop, a $90 million raise from a company reporting $60 to $80 million projection in 2026 revenue — on track for a second consecutive doubling — is not a sign of fundraising difficulty. It signals a company either approaching the point where it can finance operations from revenue, or deliberately keeping its cap table tidy ahead of a public offering. Hercules Capital's presence reinforces the second interpretation.

Soni has consistently declined to commit to a specific IPO timeline. What the Series E's composition suggests is that the question is no longer abstract.

What 450-Plus Health Systems Are Actually Getting

Before explaining where Suki is going, it is worth establishing what it has already built. The company's ambient platform is live at more than 450 healthcare organizations and licensed through more than 30 health-tech partners via Suki for Partners, its SDK and API offering. The 450-plus-organization platform metrics from the Series E announcement confirm patient encounters processed grew more than 190 percent year over year in 2026. The tenfold partner revenue growth over the past two years reflects how deeply the partner licensing business has scaled.

The technology underneath those numbers is a four-stage pipeline. A far-field microphone — on the clinician's mobile device or a dedicated room-level unit — captures the patient encounter in real time. An automatic speech recognition system with speaker diarization transcribes the audio, distinguishing clinician from patient voice and processing medical vocabulary that general-purpose speech models routinely mishandle.

The transcript then passes through a large language model that converts conversational content into a structured clinical note in SOAP or History & Physical format, coded to the EHR's requirements and calibrated to support accurate evaluation-and-management billing levels. The finished note is pushed directly into the physician's EHR via FHIR R4 or HL7 integration, with EHR integration platform details covering Epic, Oracle Health (formerly Cerner), athenahealth, and MEDITECH Expanse.

That final step — the EHR sync — is where Suki's breadth becomes a competitive differentiator. According to 2025 acute care EHR share data from Fierce Healthcare, Epic holds roughly 44 percent of U.S. acute care hospital market share; Oracle Health holds around 22 percent; MEDITECH and athenahealth serve significant ambulatory and community hospital segments. Building and maintaining certified integrations across all of them is a meaningful engineering investment, and a company that already has those integrations does not need to rebuild them from scratch for its next product layer.

An AAFP Innovation Lab study results found that Suki's assistant reduces documentation time by approximately 72 percent on average. Separately, KLAS monthly provider benefit data shows that health systems deploying Suki see an average monthly financial benefit of more than $2,600 per provider.

From Scribe to Agent: What the Pivot Actually Means

Documentation AI and agentic AI are not the same thing, and the distinction matters more in healthcare than in most other industries.

An ambient scribe produces an output — a clinical note — that a physician reviews and signs before it becomes part of the medical record or drives any clinical action. The physician's signature is the accountability checkpoint. If the AI made an error, the physician had an opportunity to catch it.

An agentic system is different. When Suki's proposed agentic layer interprets a finalized note and autonomously triggers a referral, schedules an appointment, or submits a prior authorization to a payer, those actions may happen without a pre-execution human review. A referral sent is a referral sent. A prior authorization submitted starts a payer clock. Neither waits for a physician to sign off.

"The bigger picture is that healthcare AI is moving beyond documentation," Soni said. The company plans to allocate the Series E capital across three priorities: advancing its AI models, expanding its commercial organization, and scaling its partner ecosystem.

The commercial logic is strong. Healthcare checks the structural boxes that make agentic AI valuable: highly repetitive tasks (every diabetes patient needs a referral follow-up checklist), established decision rules (prior authorization criteria are defined by payer contracts), enormous labor costs (physicians spend roughly two hours on administrative tasks for every one hour of direct patient care per the AMA 2025 prior auth survey), and a worsening workforce shortage (the AAMC projects an AAMC 86,000-physician shortage projection by 2036).

The liability question is less resolved. The FDA Software as Medical Device framework and current HIPAA guidance do not clearly assign accountability when an AI agent executes a clinically wrong downstream action. A referral generated from an AI note that contained an omission — and peer-reviewed research suggests omissions are both common and harder for physicians to detect than other error types — creates a liability chain that healthcare legal and compliance teams have not yet standardized. Suki has not publicly addressed this regulatory gap; neither, for that matter, has any ambient AI vendor at the agentic stage.

What the Evidence Says About Accuracy

Ambient AI documentation accuracy is now subject to peer-reviewed scrutiny, and the findings are not uniformly reassuring.

A January 2025 study published in the Journal of Medical Internet Research by researchers at MedStar Health and Georgetown University Medical Center examined 44 patient notes generated by AI scribes across 10 physicians and found 127 total errors — meaning 70 percent of notes contained at least one mistake. The Biro et al JMIR 2025 study found the most common error type was omissions: clinically relevant content that was present in the recorded conversation but absent from the generated note. The authors noted that omissions were also the hardest error category for physicians to detect during review—because a physician checking a note looks at what is there, not necessarily what should be there but is missing.

The relevance to Suki's agentic pivot is direct. If an autonomous agent triggers a referral or prior authorization based on AI-generated note content, and that note omits a relevant clinical fact, the downstream action may be wrong in a way that is difficult to catch before it reaches the patient or payer. Suki has not publicly named the specific human-in-the-loop review architecture it plans to apply to agentic outputs — a question that will become more urgent as the company scales this capability.

Suki Ask: Clinical Reasoning Enters the Encounter

Five days before the Series E was announced, Suki launched Suki Ask, a clinical decision-support feature built on the Wolters Kluwer UpToDate Expert AI API. The Suki Ask UpToDate Expert API launch brings evidence-based clinical reasoning into the encounter workflow — not as a separate application the clinician opens in another tab, but embedded within the ambient documentation experience.

Ask is, by Suki's account, the first UpToDate Expert AI integration into an ambient documentation platform. In early company evaluation data, 83 percent of clinicians preferred Ask over their current workflow for clinical guidance, with statistical significance (p<0.001). That figure comes from Suki's own evaluation and has not been independently replicated.

The launch extends Suki's concept of Ambient Clinical Intelligence from a documentation tool into a clinical reasoning assistant — a qualitatively different category, with different reimbursement implications and, potentially, different FDA classification exposure.

Deployments and Momentum

Several large-scale deployments marked Suki's 2026 growth trajectory. The Austin Regional 40-location Suki deployment rolled out across all of Austin Regional Clinic's Central Texas locations in July 2026, one of the larger ambulatory deployments in the company's history. The Rush University Suki expansion announcement expanded Suki's footprint at Rush University System for Health earlier in the year. More than 20 enterprise health systems have incorporated Suki's MEDITECH Expanse Suki integration launch since that partnership launched in late 2024, with expanded support for outpatient, emergency department, and inpatient settings and plans to extend ambient support to nurses and home-care staff.

In September 2026, Frost & Sullivan recognized Suki with its Frost Sullivan Innovation Recognition 2026 — the 2026 North American Transformational Innovation Leadership Recognition in Ambient Clinical Intelligence — citing documentation quality, measurable customer outcomes, and product vision.

Crowded Market, Clearer Differentiation

The competitive field for ambient clinical AI has attracted several well-capitalized players. Nuance DAX's Microsoft ambient platform retains its incumbent position, with health-system relationships built over two decades and the distribution leverage of Microsoft 365. Abridge, which received Abridge Best in KLAS recognition in both 2025 and 2026, has pursued an academic medical center strategy anchored by deployments at UPMC and similar systems, with its 2025 $300M Series E at $5.3 billion the largest financing in the category to date. Ambience Healthcare and DeepScribe round out the better-funded tier.

What has distinguished Suki within this cohort is a combination of EHR breadth (certified integrations across more than a dozen major platforms), a platform licensing business (Suki for Partners developer portal) that Soni has projected will grow to more than 50 percent of total revenue, and — increasingly — a credible agentic roadmap that competitors have not yet publicly matched in product specificity.

The agentic pivot is, in part, a pre-emptive competitive move. If ambient documentation becomes a commodity feature — available directly from EHR vendors, or from any of several AI platforms at minimal marginal cost — the companies that have already built the EHR write-access and workflow integration required for multi-step agentic action will be better positioned to retain enterprise contracts. That connective tissue is expensive and slow to build. Suki has a two-year head start on scaling it.

Whether that head start survives the accuracy question, the regulatory gap, and the competition from Microsoft-backed incumbents will determine whether Suki's next fundraise is a Series F or a prospectus.

Frequently Asked Questions

What is Suki's agentic AI, and how is it different from the ambient scribe it already sells?

Suki's current product is an ambient scribe: it listens to a patient-clinician encounter and generates a structured clinical note that the physician reviews and signs. Agentic AI takes that note as input and acts on it autonomously — triggering a referral, scheduling a follow-up appointment, or submitting a prior authorization to a payer — without requiring any additional physician action. The key difference is that the agentic layer may execute actions before a human reviews them, which raises accountability questions that ambient scribing did not. No regulatory framework currently specifies who is liable when an autonomous AI agent triggers a clinical action that turns out to be wrong.

How accurate are AI-generated clinical notes, and does that accuracy matter for the agentic layer?

A January 2025 peer-reviewed study from MedStar Health and Georgetown University Medical Center found that 70 percent of AI-generated clinical notes contained at least one error, with omissions being both the most common error type and the hardest for reviewing physicians to catch. This MedStar Georgetown ambient AI study was published in the Journal of Medical Internet Research. The accuracy gap matters directly for Suki's agentic pivot: if an autonomous agent triggers a referral or prior authorization based on a note with a clinically significant omission, the downstream action may be wrong in a way no human reviewed before it reached the patient or payer.

How does Suki compare to Nuance DAX and Abridge?

All three companies are competing in ambient clinical documentation and have significant health-system deployments. Nuance DAX Copilot benefits from Microsoft's distribution network and long-standing health-system relationships. Abridge raised $300 million in a 2025 Series E at a $5.3 billion valuation and holds a strong academic medical center position, with Best in KLAS recognition in 2025 and 2026. Suki differentiates on EHR breadth (12-plus certified integrations versus competitors' narrower lists), its Suki for Partners licensing business (which turns other health-tech platforms into distribution channels), and its explicit agentic product roadmap — the area where the next competitive battle in healthcare AI is most likely to be decided.

What is a prior authorization, and why does automating it matter?

Prior authorization is an administrative requirement by which a physician must obtain approval from a patient's insurance company before providing a specific treatment, medication, or procedure. The process consumes an estimated 13 physician-hours per week on average, according to the AMA prior authorization time burden, and is a primary driver of physician administrative burden and care delays. Automating prior authorization — generating the request, attaching the relevant clinical justification, and submitting it to the payer — is one of the highest-value workflow automation targets in healthcare. A federal rule from the Centers for Medicare & Medicaid Services, effective January 2027, requires payers to implement standardized APIs for prior authorization processing, lowering the technical barrier for AI systems like Suki's agentic layer to submit PA requests programmatically. See the CMS prior authorization API rule for full details on the January 2027 requirement.