Suki is listening - so your doctor doesn't have to type.

Somewhere, right now, a doctor is making eye contact.

It is 4:47 p.m. on a Tuesday. The internist has eighteen patients on her schedule, four refill requests, and a chart that closed at midnight. She walks into Exam Room 3 and, for the first time today, she puts her laptop down. A small app on her phone is already running. It listens. It will write the note. She will sign it.

That small app is called Suki. It is not the loudest company in healthcare AI - it does not need to be. It is the one that sits, quietly, in roughly 350 American health systems and tries to disappear. Founder Punit Soni's stated design goal for the product is, almost perversely, that you should never notice it. The pitch is short: less typing, more patients, fewer burned-out clinicians.

The reason this matters - the reason a quietly-listening voice model is something to write about at all - is that medicine in the United States has, for the better part of a decade, been losing the war against its own paperwork. Studies from the AMA and others routinely find clinicians spending more time on documentation than on patients. Burnout rates float around half the profession. The most expensive humans in the economy are using their most expensive hours to copy-paste.

Suki, the company, exists because of that exact mismatch.

The note is the enemy.

Every patient encounter in modern American healthcare produces a clinical note, which then produces codes, which produce a bill, which produces an audit, which produces - somewhere down the chain - the doctor staring at a screen at 11 p.m. The note is the source of revenue. The note is also the reason the average primary care physician spends almost two hours on the EHR for every one hour with a patient.

Punit Soni had no medical training when he noticed this. He had spent years at Google running product for Motorola - he helped ship the Moto X, one of the first phones with a real voice assistant - and then a stretch as Chief Product Officer at Flipkart in India. He liked voice. He liked big, ugly enterprise problems. In 2016 he started shadowing doctors. The thing that struck him was not a single broken feature; it was the cumulative weight of small frictions.

The answer the industry had been offering for thirty years was the medical scribe - a human, often a pre-med student, who sat in the corner of the room and typed. Scribes work. Scribes also cost roughly $30,000 a year, do not scale, and quit. The dictation incumbents (the Nuances and M*Modals of the world) had taken the problem halfway there with speech-to-text. Suki's bet was that the rest of the way was not a transcription problem at all. It was an understanding problem.

An ex-phone guy decided doctors deserved a voice assistant too.

Soni teamed up with Karthik Rajan, an engineer with a deep speech-and-systems background, and started the company in 2017 under the legal name Robin Healthcare Inc. The first version of the product was, by Soni's own retelling, an embarrassment. It could parse simple commands - "show me the last MRI" - but it could not do the thing the team actually wanted, which was to write the note itself.

There is a slightly absurd quality to the early Suki story. A team of ex-consumer-tech people, in a co-working space in Redwood City, were chasing a problem that the largest enterprise vendors in healthcare had been gnawing at for decades. They had no obvious right to win.

What they did have was a generational shift coming. Large language models, in the run-up to and aftermath of ChatGPT, would do for clinical understanding what the smartphone had done for cameras: collapse a long, hard pipeline into a single, embarrassingly capable component. Suki had been building on transformer-based speech and language models well before they were fashionable. When the wave hit, they were upright on the board.

LEADERSHIP - FOR FULL EXECUTIVE ROSTER, SEE SUKI.AI/ABOUT.

It listens. It writes. It codes. It shuts up when asked.

The current Suki Assistant does roughly four things, in roughly this order. It listens, ambiently, to the conversation between clinician and patient. It produces a structured clinical note - SOAP format, specialty-aware - written directly into the EHR. It suggests ICD-10 and HCC codes for the visit, which is where a meaningful chunk of the ROI math lives. And it answers spoken questions: “what was her last A1C?”, “summarize the cardiology consult.”

What separates Suki from the dozens of newer ambient-scribe startups is, mostly, plumbing. Building an LLM that produces a plausible note is a weekend project in 2026. Building one that round-trips into Epic, files structured orders, hands off the discrete data fields a coder needs, and survives an enterprise security review - that is a multi-year integration story. Suki has eight years of it.

By the numbers

The partners list reads like a who's-who of the EHR cartel.

In 2024, Suki became the first ambient AI to ship deep, real-time integrations with all four of the dominant US EHRs - Epic (which owns roughly a third of the hospital market), Oracle Health (formerly Cerner), athenahealth, and MEDITECH. This is not flashy. It is also the single largest moat in healthcare software, and most well-funded competitors do not have it.

On top of the EHRs sit the strategic partners. Zoom, in 2024, both invested in Suki and folded its notes into the Zoom telehealth product. Amwell embedded Suki inside its Converge platform, which is used by 55+ health plans. Premier - a hospital group-purchasing organization that buys for 4,350 American hospitals - signed a national agreement that effectively short-circuits Suki's enterprise sales cycle.

PARTNER ECOSYSTEM - PARTIAL LIST, BECAUSE THE FULL ONE GETS TIRESOME.

"Make doctors less miserable" is a mission statement disguised as a product roadmap.

Suki's official mission is to lift the administrative burden from healthcare so clinicians can focus on patients. In practice the team talks about this in more domestic terms - dinner-with-family, in-bed-by-eleven, charts-closed-by-six. The marketing copy is about “ambient clinical intelligence.” The internal language is about getting people home.

It is also a deliberately narrow mission. Suki does not, today, do diagnostic AI. It does not read radiology images. It does not recommend treatment. It just writes the note, runs the codes, and answers questions about the chart. That restraint is part of the company's actual strategy. Building a clinical co-pilot that proposes diagnoses is a regulatory and liability rabbit hole. Building one that writes notes faster than a human scribe is, by comparison, a sales conversation.

If ambient AI becomes the new clinical UI, Suki is in a position to be the OS for it.

The endgame Soni talks about, in his calmer interviews, is not really “a better scribe.” It is replacing the keyboard-and-mouse model of medicine with a voice-and-ambient one - and selling the platform underneath that shift to anyone who wants to build on top of it. That is what Suki Platform is for. It is also why the EHR integrations matter so much. The company is positioning itself less as an app and more as the connective tissue between speech, model, and chart.

Whether Suki ends up as the platform, gets absorbed into one of the EHR giants, or finds itself competing with Microsoft and a half-dozen better-funded entrants, the underlying bet has already partly paid off: ambient documentation is no longer an experimental category. It is the way meaningful chunks of American medicine now run. Suki helped make it that way.

Back in Exam Room 3, it is now 4:53. Six minutes after the visit started. The internist is walking out. The note is, somehow, already written. The patient is reaching for her coat. The internist still has seventeen people to see today, but for the first time in a long time, she is leaving the room before her note does.