Mike Ng co-founded Ambience Healthcare and now serves as its president and chairman, not its chief executive. Ambience announced in September 2025 that co-founder Nikhil Buduma had become CEO. That leadership change is essential to a current profile and provides a cleaner starting point than an unverifiable personal medical origin story.

Ambience sells AI software that listens to clinical encounters with consent and drafts documentation and related workflow outputs for clinician review. Public health-system evidence shows production adoption at Cleveland Clinic. It does not show that the product can diagnose, treat, or safely operate without professional oversight.

Current role and company timeline

Ambience’s September 22, 2025 leadership announcement says Buduma became CEO while Ng moved to president and chairman. The company also stated that it served more than 40 healthcare organizations and employed more than 200 people at that time. Those scale figures are vendor disclosures, not an independently audited market census.

A syndicated copy of the September 2025 announcement also identifies Ng as Ambience president and chairman. It preserves the role wording outside the company’s own domain, but remains a press-release syndication rather than an independent financial or clinical audit.

The change in roles need not imply a departure or dispute. The announcement says Ng remained in executive and governance positions. Public sources reviewed for this article do not disclose the board process, division of operating responsibilities, compensation, or private reasons for the change. Those gaps should remain gaps.

What Ambience offers

Ambience describes a platform that turns a patient-clinician conversation into a draft note and supports adjacent coding and clinical-documentation workflows. The company has expanded its own product language from an ambient scribe toward an operating layer for clinical work. That broader label is a vendor position, not an established category.

It helps to separate the functions:

  1. Audio capture records the clinical conversation after the required notice or consent.
  2. Speech and language systems convert the encounter into structured information.
  3. A generative system drafts a note or other proposed output.
  4. The clinician reviews, edits, and approves what enters the health record.
  5. Downstream systems may use the approved record for care, coding, billing, or quality work.

An error can enter at any step. A speaker can be misidentified. A negation can disappear. A family member’s history can be assigned to the patient. A plausible assessment can be added even though it was never discussed. If downstream automation treats a draft as verified data, one documentation error can propagate.

For that reason, “ambient” should describe the user interaction, not the control model. The software may recede from view during the visit, but consent, review, provenance, and accountability must remain visible.

Financing and company claims

Ambience announced a $243 million Series C in July 2025. The company described the round as supporting product development and health-system expansion. Press accounts attached a private valuation to the deal, but the official source cited here does not provide audited financial statements or enough detail to value the company independently.

The round confirms that investors supplied the announced amount on the company’s stated timeline. It does not verify revenue, margin, retention, customer count, or claimed improvements in coding and clinician productivity. Ambience is private, and its public materials do not expose cohort-level operating data.

The correct approach is to attribute deployment or performance figures to the organization that measured them, identify the population and period, and preserve study limitations. A percentage without a baseline, denominator, comparator, and method is not decision-grade evidence.

Cleveland Clinic offers a verifiable deployment record

Cleveland Clinic announced in February 2025 that it was rolling out Ambience across more than 80 specialties. Its release says patients are notified and can opt out, clinicians review and approve notes, and the tool does not diagnose or treat patients.

A later Cleveland Clinic implementation account says more than 4,000 clinicians used the system within 15 weeks. It also describes verbal consent and clinician review. This is evidence from a named deploying health system rather than Ambience alone, but it is still the institution’s account of its own program. It does not provide randomized evidence or a complete error distribution.

The distinction matters. A rollout demonstrates that a large provider integrated and governed the tool. It does not establish that every note is more accurate, that clinician burnout declined for every specialty, or that the product improved patient outcomes. Those require studies designed for the specific question.

Independent research supports cautious measurement

A 2025 JAMA Network Open study followed 46 clinicians using an ambient documentation tool. It found associations with less note time and after-hours work, along with mixed qualitative feedback. The study was a single-group pre-post quality-improvement evaluation, not a randomized trial, and it did not test Ambience specifically.

That evidence supports the plausibility of reducing documentation burden. It does not validate a particular vendor. More recent simulation research across specialties found differences in documentation quality and noted that rigorous independent comparisons remain scarce. Again, the vendors were not identified, so the study belongs in the category evidence base rather than Ambience’s scorecard.

The combined lesson is useful: ambient AI can change clinician workflow, but quality varies by specialty, scenario, and product. Adoption should include prospective monitoring rather than assume that an encouraging average transfers to every clinic.

A clinical evaluation protocol

A health system should evaluate proposed notes against the encounter and the source record. The sample needs common visits, specialty-specific terminology, interpreters, multiple speakers, background noise, sensitive topics, medication changes, uncertain diagnoses, and explicit negation.

Measure at least:

  • factual omissions and unsupported additions;
  • medication, allergy, dosage, and laterality errors;
  • incorrect speaker or family-history attribution;
  • editing time and material correction rate;
  • note closure and after-hours documentation time;
  • patient notice, consent, and opt-out completion;
  • performance by specialty, language, accent, and encounter type; and
  • downstream coding or billing changes after clinician approval.

Critical errors should be reported separately from stylistic edits. A single fabricated medication instruction matters more than several formatting changes. The system also needs a channel for clinicians to flag errors, with version and model information attached so recurring patterns can be investigated.

The test should compare against the current workflow, not an idealized baseline. If clinicians use templates, dictation, scribes, or copy-forward practices today, those alternatives have their own error and cost profiles.

Privacy and operating controls

Clinical audio and notes can contain protected health information. A health system needs a documented data flow: capture device, transport, storage, model provider, subprocessors, electronic health record integration, logs, backups, and deletion. Contract language should identify permitted uses, retention, breach duties, and whether any data can train a model.

The U.S. Department of Health and Human Services Trustworthy AI Playbook offers a public-sector framework for governance, risk analysis, privacy, monitoring, and accountability. It is not a certification of Ambience and does not replace HIPAA analysis. It is useful as a checklist for an institution building its own control process.

Consent must also work in practice. Patients need an understandable explanation, a usable refusal path, and equivalent care if they opt out. Clinics should decide how to handle minors, interpreters, telehealth, companions, and visits involving especially sensitive information. A consent statement in a launch release does not prove consistent execution.

Assessing Ng and Ambience

The evidence supports a focused conclusion. Ng co-founded a company that attracted substantial capital and achieved a documented large-system rollout. He remains president and chairman after Buduma became CEO. Ambience is participating in a category with credible potential to reduce documentation work.

The public record does not establish the personal injury narrative in the former article, a precise valuation from primary financial documents, service to more than 100 health systems, or a universal improvement in coding or burnout. It also does not show clinical outcome gains attributable to the product.

Ambience should be judged on the accuracy and reviewability of its drafts, consent and privacy execution, specialty performance, integration reliability, and measured reduction in total work. Ng’s current role should be judged from disclosed company governance, not reconstructed private motives.

Source and correction note

This revision uses Ambience’s dated financing and leadership disclosures, Cleveland Clinic’s deployment records, a university profile, independent clinical research, and federal governance guidance available through September 13, 2026. Vendor figures are labeled as such and category studies are not represented as Ambience trials. The earlier version used an unsupported personal medical story, an outdated CEO title, and unverified customer, valuation, and performance figures. Those claims have been removed or corrected.