Selling outcomes, not seats: two health-AI vendors put their own fees at risk on the hospital's numbers
Health AI has spent three years selling a per-clinician-per-year licence. This month two vendors said they want out of that deal. On August 19 Ambience Healthcare launched The Ambience Standard, putting its platform fee at risk against each health system's own clinical, operational and financial targets (Business Wire / Yahoo Finance, Aug 19, 2026). On August 13 Hippocratic AI shipped 30-plus Agentic Orchestrators, moving the unit of delivery from "one phone call" to "readmission rate" and "Star Rating" (PR Newswire, Aug 13, 2026). In the same month Cleveland Clinic published its four-month, 4,000-clinician rollout as a paper in npj Health Systems (Nature, Aug 11, 2026). All three point one way: once the products converge, the only thing left to sell is a willingness to be held to the result. The catch is that hospitals may not want that contract. Tampa General's innovation chief put it plainly: "We want to be able to model out what an ROI would be… then have fixed pricing" (MedCity News).
01 — Top Stories
Ambience puts its own fee on the table: miss the health system's targets, collect less
On August 19 Ambience Healthcare announced "The Ambience Standard," which it frames as a rewrite of how AI vendors contract with health systems: a dedicated team of clinicians, engineers and transformation specialists embedded per customer, and fees tied not to tokens, seats or features but to that system's own measurable targets, with the vendor's compensation at risk if they are missed. The customer figures released alongside: Cleveland Clinic onboarded 4,000 clinicians in four months at 70% utilisation and an NPS of 60; Ardent Health reported 3x validated ROI from coding improvement and time savings; Onvida Health about $24,000 of positive annual financial impact per physician; MultiCare 92% clinician adoption with an NPS 63 points above a competing product; St. Luke's saw physician burnout fall from 45% to 31% and intent-to-leave from 31% to 18% (Business Wire / Yahoo Finance, Aug 19, 2026; HIT Consultant, Aug 19, 2026).
Product differentiation in ambient documentation has flattened out — Endpoints News called the sector a "great convergence" in early August, arguing the fight has shifted from capability to share (Endpoints News, Aug 4, 2026). In that setting, "my fee rides on your numbers" is one of the few remaining levers — and it is a lever at the contract layer, so a rival who copies it has to put gross margin behind it too. Co-founder Mike Ng's line lands: "There's a million miles between a better model and better care."
All five customer figures come from the vendor's own release with no independent audit; neither the denominator behind "3x ROI" (contract value) nor the method behind "$24,000 per physician" is disclosed. More to the point, the release never says what share of the fee is actually at risk — 5% or 40% decides whether this is marketing or genuine risk-sharing.
Hippocratic AI bundles voice agents into "orchestrators": the unit of delivery moves from task to outcome
On August 13 Hippocratic AI introduced Agentic Orchestrators: an orchestration layer supervising teams of specialised voice agents, deciding which agent reaches which patient, when and how. The first wave of 30-plus orchestrators splits three ways — payer (chronic care management, readmission prevention, Star Ratings), provider (lost-to-follow-up recovery, inpatient and ambulatory care, leakage reduction) and life sciences (drug launches, trial enrolment, adherence, real-world evidence). Underneath sits the company's Polaris architecture: a 700-billion-parameter primary model plus 30-odd supervising models. Hippocratic reports 99.89% correct advice with zero severe-harm events, validated by 7,700-plus US-licensed clinicians across 775,000 calls, on a base of 250 million completed clinical patient interactions and 300-plus live clinical use cases (PR Newswire, Aug 13, 2026; Hippocratic AI newsroom).
This is the same move as Ambience's, written differently. Ambience changes the contract; Hippocratic changes the packaging unit — from "here is an agent that makes calls" to "here is a programme that lowers readmissions." Both rest on the same premise: buyers no longer want to assemble tools into results themselves. Bessemer named the pattern in January's State of Health AI 2026 — "AI-services-as-software," delivering service-level outcomes at 70–80%-plus software margins (Bessemer Venture Partners, Jan 2026).
The 99.89% and the "zero severe harm" claim are vendor-reported: the release gives no evaluation protocol, no definition of harm, and no evidence the reporting channel is independent. Note too that the announcement names no customers. A product built on accountability for outcomes that launches without one buyer willing to be named is itself a data point.
4,800 clinicians, 3.5m encounters: Cleveland Clinic turns "how to deploy" into a paper
On August 11 npj Health Systems published a paper co-authored by Cleveland Clinic and Ambience Healthcare setting out an enterprise-scale ambient-documentation deployment as four pillars: governance, training and onboarding, support, and real-time learning. Governance ran through an Executive Sponsor Group and a Project Operating Council, with cross-functional teams covering IT, revenue cycle, training, product feedback and adoption monitoring. The numbers: 4,000-plus ambulatory clinicians onboarded within four months; 4,800-plus using the system across more than 3.5 million encounters within twelve; among established users (50-plus uses) 70% encounter-level utilisation, NPS 60, CSAT 96.6%, and 60% agreeing the tool increased their likelihood of staying in practice (npj Health Systems, Aug 11, 2026).
This daily covered the mirror case on August 24: a record-reading AI quietly abandoned by clinicians. This paper answers that gap — it does not argue the model is good, it argues that the deployment method is itself a reproducible asset. Which also explains why Ambience is willing to put fees at risk: if 70% utilisation is manufactured by process rather than luck, the risk of promising an outcome becomes quantifiable.
The paper is co-authored with the vendor — the single most important caveat. The authors state that Cleveland Clinic operational data are not publicly available, so nothing here is externally verifiable; burnout and time-savings figures are cited from outside literature rather than measured in this deployment; there is no control arm and no randomisation. It is a good implementation guide, not efficacy evidence.
Hospitals actually want a fixed price: the biggest obstacle to outcome pricing is the buyer, not the tech
While vendors push toward selling outcomes, the buyer-side signal runs the other way. Tampa General Hospital's Rachel Feinman: "We want to be able to model out what an ROI would be… then have fixed pricing," adding that the system has not been receptive to revenue-sharing. Oak HC/FT's Vig Chandramouli notes that pricing off success rates, if set too low, breaks the vendor's own unit economics. Healthcare IT Leaders' Joey Dean proposes the middle path: rather than per-transaction billing, both sides define shared goals first, then contract to them (MedCity News).
The line worth keeping from this issue: the bottleneck on outcome pricing is the hospital's budgeting process, not the model. Systems set annual budgets and answer to boards; a fee that floats is a liability, not a discount. Which makes Ambience's design shrewd — it is not revenue-share but "we lose fee if targets are missed," giving the buyer a fixed ceiling with downside protection, far easier to get through procurement than genuine risk-sharing. It is also why that design is likely to be copied while Hippocratic-style pay-for-outcome spreads slowly.
The MedCity News piece carries no clearly stated publication date on the page (its URL path reads 2026/03), and three interviewees do not represent the buyer market. Treat this section as context, not as survey data.
Tempus posts its first profitable quarter — and buys MRD in for $1.5bn
Tempus AI reported Q2 2026 on July 30: revenue $382.5m, up 22% (Diagnostics $289.3m, +20%; Data & Applications $93.2m, +28%, with Insights +36%), gross profit $246.5m (+26%), and net income of $5.6m — against a $42.8m net loss in Q2 2025, this is the quarter it crossed over. Adjusted EBITDA was $8.0m; cash and marketable securities $820.7m. Full-year guidance was raised to $1.595–1.605bn (about 25% growth) with roughly $65m adjusted EBITDA. In the same quarter MRD volume rose from 6,500 to 9,000 tests, the xT Tumor Only assay won FDA approval, the first oncology foundation model was delivered to AstraZeneca, and roughly $200m of data-licensing bookings were signed (Tempus press release, Jul 30, 2026). On July 20 it announced the acquisition of Personalis at about $1.5bn enterprise value, folding molecular residual disease into its precision-oncology platform; Personalis shares were still moving on deal progress in late August (Tempus, Jul 20, 2026; MedTech Dive).
Tempus is the other answer to this issue's question. Where Ambience and Hippocratic differentiate by contracting for outcomes, Tempus differentiates by owning the input: assay volume, sequencing data, pharma licences. Roughly $200m of data-licensing bookings in a single quarter, plus an oncology foundation model handed to AstraZeneca, says what it actually sells is training data nobody else can assemble. Two very different moat shapes — one built on a promise, one on a stock.
A $5.6m profit on $382.5m of revenue is a 1.5% net margin, booked in the same quarter as a $460m convertible note offering and a $1.5bn acquisition announcement. The Personalis deal has not closed, so neither integration cost nor dilution appears in these numbers. "First profitable quarter" is accurate; "durably profitable" is not yet.
Epic's Agent Factory is the biggest variable in this pricing experiment
This daily covered Epic UGM 2026 in full on August 21; only three items matter to today's thread. Agent Factory: a no-code platform to build, configure and monitor AI agents, with 120 pre-built capabilities; early adopter ECU Health saved roughly 20 hours a week on transfer-centre summaries; wide availability 2027. Curiosity: a generative model trained on Cosmos — 320m-plus patients, 23bn encounters — predicting readmission and stroke risk, with 20 organisations validating and a March 2027 release. Chart with Art: native ambient documentation, now spanning 70-plus specialties. Epic reports 43.7% acute-care EHR share, 3,700-plus hospitals, 325m-plus patient records, and 85%-plus of customers using Epic AI as of March 2026 (Fierce Healthcare, Aug 19, 2026; TechTarget, Aug 18, 2026).
What Epic sells is already inside the contract. Once a system's EHR licence includes Chart with Art and 120 agent capabilities, an independent vendor asking procurement for a separate line item has to offer something Epic cannot — which is exactly what "my fee rides on your KPI" is for. The pricing pivot at Ambience and Hippocratic reads less as business-model innovation than as a defence against bundling. Epic international president Carl Dvorak's UGM line — "we've got to find a way to move faster" — is bad news for the independents.
This week's cheques: Happy Health $75m, MaxQ Medical $31.5m, Hopscotch $53m
Between August 18 and 20 Fierce Healthcare's tracker logged four rounds. Happy Health raised a $75m Series A (ARCH Venture Partners and OpenLoop leading) for AI-powered continuous home-based care, entering through an FDA-cleared clinical-grade smart ring aimed first at sleep. Hopscotch raised $53m Series D (8VC, Townhall Ventures) for technology-enabled rural primary care, running 12 clinics in western North Carolina and having served over 15,000 patients. MaxQ Medical took $31.5m Series A for an integrated transurethral imaging-and-therapy system, starting with benign prostatic hyperplasia. Ours Privacy raised $15m Series A for HIPAA-compliant marketing data infrastructure, already used by 200-plus providers and payers (Fierce Healthcare Fundraising Tracker '26). For contrast, July's largest were Function Health's $450m growth round and Neko Health's $700m Series C (New Market Pitch, Aug 2026).
Four rounds, roughly $174m combined, and not one of them in clinical-decision AI. The money went two ways: hardware entry points (a smart ring, an imaging-plus-therapy system) and administrative infrastructure (a compliance data layer). That matches this issue's thread — as pure-software AI features get absorbed into platforms like Epic, what can still be priced on its own either has a physical form or sits inside a compliance workflow. Bessemer counted AI companies taking 55% of health-tech funding in 2025, up from 37% in 2024 — but the definition of "AI company" is drifting downstream (Bessemer, Jan 2026).
A fundraising tracker records only publicly announced deals; undisclosed or unannounced rounds are absent, and a single week is too small a sample to support a trend claim. Read the interpretation above as an observation, not a conclusion.
Taiwan's FHIR Box goes live: medical centres interoperable by end-2026, clinics by 2028
On August 24 Taiwan's Ministry of Health and Welfare launched "FHIR Box," positioned as a universal operating layer for healthcare data, intended to make medical records interoperable across hospitals on a common standard. The timeline: record interoperability at all major medical centres by end-2026, extension to regional and district hospitals by end-2027, and clinics and health stations by 2028. Named beneficiaries include Long Chia Intelligent (6841 — H1 2026 gross margin up from 57.8% to 73.1%, 57 medical-device certifications including 15 FDA), Acer Smart Healthcare (6857 — H1 revenue up 38.4% year on year), Far EasTone (4904, systems integrator for the nationwide hospital upgrade), plus Jinn Horng (6796) and Cheer Tech (8409). On the government side, three national AI centres handle ethical standards, multi-centre clinical validation using federated learning, and reimbursement assessment; 44 projects are complete, 9 have regulatory approval and 13 are pending (Yam News, Aug 24, 2026; MOHW Taiwan Smart Healthcare Centres).
Outcome-based pricing has a precondition: outcomes must be measurable, and measurement needs data comparable across institutions. That is precisely what Taiwan has lacked — every hospital's format differs, so "this system cut readmissions by N points" could not be verified across organisations. If FHIR Box actually lands, it removes not only a technical barrier but the prerequisite for Taiwanese vendors to write American-style outcome contracts at all.
This item comes from financial-press coverage rather than the ministry's own release, and the margin and revenue figures for individual listed companies are as relayed by that report rather than checked line by line against Taiwan's official filings database. Naming beneficiary stocks is not investment advice.
02 — Product Analysis
The Ambience Standard
Outcome-linked deployment contract · Ambience Healthcare (US)
Function and position. Not a new model but a bundle sold as one contract to health systems: ambient documentation platform, embedded deployment team, and outcome-linked pricing. The buyer is a large system capable of defining measurable targets — Cleveland Clinic, Ardent Health, MultiCare, St. Luke's, Onvida Health — and the pitch is that some of the implementation risk moves off the hospital and back onto the vendor (Business Wire, Aug 19, 2026).
- Strength : the deployment method is now a peer-reviewed paper — four pillars, a two-tier governance structure, 4,000 clinicians in four months, 70% encounter-level utilisation. That is organisational capability a rival cannot simply buy (npj Health Systems, Aug 11, 2026).
- Strength : on share, Ambience sits third at about 13%, behind Microsoft/Nuance at 33% and Abridge at 30% (Menlo Ventures survey, Becker's Hospital Review). Third place has far more incentive than first to trade margin for contracts — which makes this pricing move a sensible bet for it specifically.
- Concern : the share of fee genuinely at risk is not disclosed. Without that number, "fees at risk" and "a rebate on hitting target" may be economically identical. Also undisclosed: who adjudicates whether a target was met, and how disputes are settled.
- Concern : an embedded team is a labour-intensive delivery model and compresses margin. Bessemer argues the category can hold 70–80% software margins (Bessemer, Jan 2026), but that is a claim about the category — Ambience's own gross margin is not public.
Hippocratic AI Agentic Orchestrators
Outcome-shaped voice-agent fleets · Hippocratic AI (US)
Function and position. No longer single-purpose agents but a fleet under an orchestration layer, each fleet mapped to one concrete outcome: fewer readmissions, a better Star Rating, lost-to-follow-up patients recovered. Buyers span payer, provider and life sciences, roughly ten orchestrators each, with AI Front Door, Nurse Co-Pilot and AI Call Supervisor sold alongside (PR Newswire, Aug 13, 2026).
- Strength : the Polaris split — a 700bn-parameter primary model plus 30-odd dedicated supervisors — is a defensible engineering choice for clinical voice: safety checks live in separate models rather than depending on the primary model to police itself. And 250 million completed clinical interactions is a leading scale in this category (Hippocratic AI).
- Strength : shaping the product around a metric that already carries a dollar value — Star Ratings — is smart positioning. Payers can price each rating point precisely, which gives "selling outcomes" a ready-made basis in procurement, far easier than selling the same idea to a hospital.
- Concern : not one named customer on launch day. The "300-plus live clinical use cases" is a cumulative company figure, not adoption of these 30-plus orchestrators. What a product line sold on outcomes most needs is exactly what is missing: who deployed it, and what happened.
- Concern : the 99.89% accuracy and zero-severe-harm claims are self-evaluated and self-published. The technical scorecards this daily reviewed on August 27 all failed on the same axis — whether the process can be audited — and there is no auditable evaluation protocol or third-party replication path here either.
03 — Companies & Competition
| Company | Recent state & numbers | Position & moat |
|---|---|---|
| Ambience Healthcare Third in ambient scribing, attacking through contract design |
About 13% share, behind Microsoft/Nuance at 33% and Abridge at 30% (Becker's / Menlo Ventures). Launched outcome-linked contracting Aug 19; Cleveland Clinic 4,000 clinicians in four months, Ardent Health 3x ROI, Onvida about $24k per physician, MultiCare 92% adoption (Business Wire, Aug 19, 2026). Ardent has separately announced enterprise-wide rollout past one million patient encounters (Ardent Health). | The moat is implementation, not the model, and it has been hardened into a teachable process by publication. Weakness: it is behind on scale, and trading margin for contracts means thinner unit economics than its rivals — if Epic's Chart with Art absorbs mid-sized systems, it has to concede more to hold ground. |
| Hippocratic AI Clinical voice agents, repackaged around outcomes |
$444m raised to date from a16z, Kleiner Perkins and NVIDIA among others; 30-plus orchestrators launched Aug 13, with claims of 250m clinical interactions, 300-plus live use cases and 99.89% correct advice (PR Newswire, Aug 13, 2026). | The moat is the safety architecture plus interaction volume — Polaris's supervisory layer and accumulated conversational data. Weakness: its core buyers are payers and pharma, and distance from the EHR is actually defensive; but with no named customers, the orchestrators' real-world effect cannot be checked from outside. |
| Abridge Leading independent, enterprise licensing |
About 30% share; roughly $100m ARR (May 2025, up from $60m at end-2024) on enterprise licences of about $2,500 per clinician per year, with 90-plus disclosed customers including Kaiser Permanente (24,600 physicians), Mayo Clinic, Johns Hopkins, Duke and UPMC; $5.3bn valuation, with a $316m Series E extension in April 2026 (Sacra; Fierce Healthcare). | The moat is procurement inertia: consecutive Best in KLAS awards and a marquee logo list make it the safe answer in an RFP. Its weakness is this issue's thesis — the per-clinician licence is precisely the surface under attack, and if outcome contracts become standard, the leader's price premium is the hardest to defend. |
| Microsoft / Nuance Share leader, distribution and bundling |
About 33% share, the single largest vendor; DAX priced at roughly $600 per month (Becker's / Menlo Ventures; Sacra comparison). | The moat is enterprise relationships and cloud bundling — the buyer usually already has a Microsoft agreement. Weakness: a large vendor's cost structure is the worst fit for an outcome-linked model that requires embedded delivery teams. |
| Epic The platform, absorbing features into the licence |
43.7% acute-care EHR share, 3,700-plus hospitals, 325m-plus patient records; 85%-plus of customers on Epic AI as of March 2026; UGM 2026 brought Agent Factory (120 pre-built capabilities, wide availability 2027) and Curiosity (Cosmos: 320m patients, 23bn encounters, releasing March 2027) (Fierce Healthcare, Aug 19, 2026). | The moat is the licence already signed and the position in the workflow; Cosmos gives its predictive models training data nobody else can obtain. Weakness is timing — Agent Factory and Curiosity both land in 2027, and that gap is the only window the independents have left. |
| Tempus AI Data-asset model, moat on the input side |
Q2 2026 revenue $382.5m (+22%) with $5.6m net income against a $42.8m loss a year earlier; full-year guidance $1.595–1.605bn; roughly $200m of data-licensing bookings in the quarter and the first oncology foundation model delivered to AstraZeneca; announced the acquisition of Personalis at about $1.5bn (Tempus, Jul 30, 2026; MedTech Dive). | The moat is the stock of clinical and molecular data; buying Personalis fills in longitudinal MRD. Weakness: a net margin of about 1.5%, an unclosed acquisition, and no disclosed renewal rate on data-licensing revenue. |
| OpenEvidence Physician-facing evidence front door, free plus advertising |
Valuation doubled to $12bn in January 2026 on a $250m Series D (CNBC, Jan 21, 2026; Fierce Healthcare), after a $210m round at $3.5bn in July 2025 alongside the DeepConsult agent launch (OpenEvidence). | The moat is distribution that bypasses procurement — it reaches the individual physician without convincing a CIO. It is the only company here that never has to answer the outcome-contract question, because it does not bill hospitals at all; the flip side is that its revenue quality rides entirely on the advertising market. |
| OpenAI General-model vendor, entering via the enterprise workspace |
Launched ChatGPT for Healthcare on January 8, 2026 — a GPT-5-based enterprise workspace with templates for discharge summaries, patient instructions, clinical letters and prior authorisation, HIPAA handling and BAAs. Early customers include Boston Children's, Cedars-Sinai, Stanford Medicine Children's Health, AdventHealth, HCA Healthcare, Baylor Scott & White, MSKCC and UCSF (Fierce Healthcare, Jan 8, 2026; OpenAI). | The moat is the model and the brand, but its position in healthcare is the most awkward: coverage notes its EHR interoperability remains unclear, and without a place in the workflow it cannot be accountable for an outcome. It is the one player here that structurally cannot price on results. |
Today's competitive structure is a squeeze. Above sits Epic, folding features into a licence already signed; below sits distribution like OpenEvidence's, reaching physicians without ever passing procurement. Caught between them is every independent living on a per-clinician-per-year fee. What Ambience and Hippocratic did this month is the same escape route — a retreat from selling capability to selling accountability — because accountability is the thing a platform finds inconvenient to take on and a distribution play cannot take on at all. Whether the route works is not theirs to decide; it is decided on the hospital's budget sheet.
04 — Taiwan Angle
(1) FHIR Box is the precondition for any of this in Taiwan. An outcome-linked contract requires measuring what changed after deployment — readmissions, coding completeness, clinician retention — and measurement requires a data format comparable across institutions. The FHIR Box the ministry launched on August 24 lays exactly that layer, running from medical centres by end-2026 to regional and district hospitals by end-2027 and clinics and health stations by 2028 (Yam News, Aug 24, 2026). Until then a Taiwanese hospital that wanted an outcome contract could not produce a baseline both sides would accept.
(2) A single-payer system makes "outcome" simpler to define than in the US — and riskier. Hippocratic ties orchestrators to Star Ratings because American payers can convert each rating point into money. Taiwan has no Star Ratings, but it has National Health Insurance reimbursement and quality indicators — and one of the ministry's three national AI centres exists precisely to handle reimbursement assessment (MOHW Taiwan Smart Healthcare Centres). Pricing power over outcomes therefore sits with one institution: the upside is that inclusion in reimbursement scales nationally at once; the downside is that a vendor has nowhere to spread its bets, and a single assessment can decide a whole product line.
(3) What this means concretely for Taiwanese vendors: build the deployment method before the model. The most instructive thing in this issue is not Ambience's contract but the content of the Cleveland Clinic paper — four pillars, two-tier governance, 4,000 clinicians in four months, 70% utilisation (npj Health Systems, Aug 11, 2026). Long Chia Intelligent lifting H1 gross margin from 57.8% to 73.1% on 57 device certifications including 15 from the FDA shows Taiwan can clear regulatory and productisation hurdles (Yam News, Aug 24, 2026). What is missing is the step that turns "how the hospital actually adopts it" into a reproducible asset — and that step, unlike the data layer, can be started today.
05 — Further Reading
-
Accelerating ambient AI scribe enterprise-scale deployment: Cleveland Clinic's novel approach to health system-industry partnership — npj Health Systems (2026-08-11)
The only peer-reviewed account that turns enterprise health-AI deployment into a reproducible process. Read the governance section rather than the utilisation figures — and note the vendor on the author list.
-
State of Health AI 2026 — Bessemer Venture Partners (2026-01)
Its "AI-services-as-software" frame explains why both companies pivoted in the same month: service-level outcomes at software-level margins. It also carries the 2025 deal count — 527 rounds, roughly $14bn.
-
AI Startups Are Tying Fees to Completed Tasks. Will Hospitals Buy In? — MedCity News
The only piece here written from the buyer's chair. A vendor release will never tell you why procurement stalls; this does.
-
Ambient AI scribes, by market share — Becker's Hospital Review（引 Menlo Ventures 調查）
One table explains why it was third-placed Ambience — not Microsoft in first or Abridge in second — that put fees at risk first. Note the survey's base date is older; treat the numbers as structural, not current.
-
Epic expands AI ambitions with agent platform, Cosmos-powered predictions and workflow automation — Fierce Healthcare (2026-08-19)
To understand the urgency among independents, look at what Epic intends to fold into the licence by 2027. Read Agent Factory's 120 pre-built capabilities as a list of things about to become free.
06 — References
- Ambience Healthcare Sets a New Standard for AI Partnerships in Healthcare. Business Wire / Yahoo Finance, 2026-08-19. finance.yahoo.com
- Ambience Healthcare Launches The Ambience Standard, Linking AI Platform Fees Directly to Measurable Clinical and Financial Outcomes. HIT Consultant, 2026-08-19. hitconsultant.net
- Hippocratic AI Announces Next Generation of Healthcare AI: Orchestrators Focused on Outcomes, Not Tasks. PR Newswire, 2026-08-13. prnewswire.com
- Hippocratic AI Press Room. Hippocratic AI. hippocraticai.com
- Accelerating ambient AI scribe enterprise-scale deployment: Cleveland Clinic's novel approach to health system-industry partnership. npj Health Systems, 2026-08-11. nature.com
- Ambient Scribes Improve Clinician Retention, Cleveland Clinic Research Shows. MedCity News, 2026-08. medcitynews.com
- AI Startups Are Tying Fees to Completed Tasks. Will Hospitals Buy In? MedCity News. medcitynews.com
- The great ambient scribe convergence. Endpoints News, 2026-08-04. endpoints.news
- Ambient AI scribes, by market share (Menlo Ventures data). Becker's Hospital Review. beckershospitalreview.com
- Ardent Health Surpasses 1 Million Patient Encounters Supported by Ambient AI. Ardent Health. ardenthealth.com
- Ardent Health plans enterprisewide rollout of Ambience Healthcare's AI platform. Fierce Healthcare. fiercehealthcare.com
- Tempus Reports Second Quarter 2026 Results. Tempus AI, 2026-07-30. tempus.com
- Tempus to Acquire Personalis, More Tightly Integrating Molecular Residual Disease (MRD) into Its AI-Enabled Precision Oncology Platform. Tempus AI, 2026-07-20. tempus.com
- Tempus to buy cancer test maker Personalis for $1.5B. MedTech Dive. medtechdive.com
- Epic expands AI ambitions with agent platform, Cosmos-powered predictions and deeper workflow automation. Fierce Healthcare, 2026-08-19. fiercehealthcare.com
- Epic unveils AI-driven Ergo Visit at 2026 UGM. TechTarget, 2026-08-18. techtarget.com
- Epic unveils AI agents, showcases new foundational models. Healthcare IT News. healthcareitnews.com
- Epic's Big Bet: Three Themes From UGM 2026. Forbes, 2026-08-25. forbes.com
- FDA promises new AI guidance, and what happened at Epic's UGM. STAT News, 2026-08-25. statnews.com
- Abridge revenue, valuation & funding. Sacra. sacra.com
- Abridge scores $300M series E, boosting valuation to $5.3B. Fierce Healthcare. fiercehealthcare.com
- OpenEvidence, the "ChatGPT for doctors," doubles valuation to $12 billion. CNBC, 2026-01-21. cnbc.com
- OpenEvidence clinches $250M Series D, rapidly growing its reach to doctors. Fierce Healthcare. fiercehealthcare.com
- OpenEvidence Announces $210 Million Round at $3.5 Billion Valuation. OpenEvidence. openevidence.com
- OpenAI rolls out ChatGPT for Healthcare, a gen AI workspace for hospitals and clinics. Fierce Healthcare, 2026-01-08. fiercehealthcare.com
- Introducing OpenAI for Healthcare. OpenAI. openai.com
- State of Health AI 2026. Bessemer Venture Partners, 2026-01. bvp.com
- Fierce Healthcare Fundraising Tracker '26. Fierce Healthcare. fiercehealthcare.com
- Digital Health Market Funding News (August 2026). New Market Pitch. newmarketpitch.com
- Large funding rounds help boost digital health investment in H1. Healthcare Dive. healthcaredive.com
- 台灣醫療AI掀全國基建！FHIR Box串聯跨院 長佳智能、遠傳受惠. 蕃新聞, 2026-08-24. n.yam.com
- 臺灣智慧醫療三大中心. 衛生福利部. aicenter.mohw.gov.tw
- 2026 年智慧醫療趨勢：從生成式 AI 到制度化治理的關鍵拐點. CIO Taiwan. cio.com.tw