Wellsheet

The models drive the core modules, and the platform carries value without them. Wellsheet's Care Team Copilot uses AI to read the chart and generate hospital course, assessment and plan and discharge summary text, to answer questions in an embedded chat agent, and to walk clinical pathways; the older product pages describe machine learning that prioritizes chart content and learns what each clinician needs.

Around that sit an EHR-embedded interface that predates the generative features, licensed UpToDate content (pathways, calculators and lab interpretations filled with the patient's data), and a discharge planning dashboard. Remove the models and the interface, content and dashboard remain.

Autonomy is claimed and oversight is asserted without a mechanism. The product generates clinical documentation, answers chart questions in the EHR, and on pathways "delivers the AI's chosen answer" with supporting evidence from the chart. The company says it reads the chart rather than listening to the visit, and a customer quote says clinicians can be confident surfaced data comes from the record.

No review step, confidence signal, threshold or escalation rule is described for generated text or pathway answers. Views adapt to each clinician's role and usage patterns, and the company does not say whether anything is suppressed rather than ranked lower, or what review is required before generated documentation is signed. The site footer says some described features are illustrative and not in the product, without saying which.

The architecture is described in general terms. The company says AI reads the whole chart to draft documentation, machine learning prioritizes chart content and learns each clinician's needs, and an agent powers chat and pathways; it holds a patent on its EHR interface. No model, model family, training data, version or update practice is named, and the footer disclaimer leaves unclear which described features ship.

The architecture is described and no model provider is named. The product is cloud and API based, reads the EHR, and fills UpToDate content from Wolters Kluwer with patient data; the privacy notice mentions unnamed cloud hosting and analytics providers. Which model generates text, whether patient data reaches Wolters Kluwer when UpToDate content is filled, and whether Elsevier or ClinicalKey AI systems now receive data are not stated.

Named customers and vendor figures with no method. The company names Ascension (system wide), RWJBarnabas Health (seven hospitals), Concord Hospital, San Juan Regional Medical Center and AllianceChicago, and Elsevier's release counts 139 hospital sites.

Its figures differ across its own material: charting time cut by 50 percent and two hours saved per clinician daily on the home page, 90 to 120 minutes a day in a February 2026 release, 90 minutes for hospitalists and a 12 to 16 percent length of stay reduction in a July 2025 release, and 40 percent less EHR time with 20 percent better throughput in an RWJBarnabas quote. None carries a period, denominator or method. Clinician counts run from 10,000+ on the home page to 65,000+ in the February release. The KLAS reports and the Ascension case study are behind forms.

General assurances. A February 2026 release calls the product HIPAA compliant end to end. The privacy notice (last modified 5 May 2026, published as a PDF on a file-sharing link) covers the website only and points app users elsewhere; nothing states retention, training use or de-identification for chart data.

Elsevier plans to combine the product with ClinicalKey AI inside its Clinical Solutions business, and nothing says whether patient data stays separate from the rest of the group or informs product development beyond the customer. No safety engineering for generated text is published.

Compliance is claimed without the document behind it. The February 2026 release calls the product "HIPAA compliant end-to-end"; no business associate status or agreement terms are published, and the privacy notice covers the website. Elsevier acquired the company on 3 June 2026, so existing agreements were signed with an entity that is now part of a larger group.

A SOC 2 mark is presented as the company's own, with the type, period and scope unstated. The home page carries a SOC 2 badge; no report, trust center, penetration testing statement or security page is published. Now that Wellsheet sits under Elsevier, nothing indicates whether the report covers Wellsheet as it runs today or the parent's environment.

No device claim is made and no regulatory position is published. FDA's 510(k) database lists no clearance for Wellsheet. The product delivers the AI's chosen answer on clinical pathways with the chart evidence beside it, fills lab interpretations and calculators with patient data, and raises discharge alerts; the company's news page lists a Newsweek story whose title promises "Virtually Zero" risk. Showing the evidence behind a recommendation fits the non-device decision support criteria, but the company has not published that analysis, and Elsevier's plans add patient engagement.

Policy language, no evaluation. The company and its new owner describe governed, trustworthy AI: Elsevier's acquisition release speaks of verified content and quality oversight, and the pathway feature grounds guidance in licensed UpToDate content rather than model output. No evaluation of Wellsheet's outputs, by patient group or otherwise, is published. Two gaps follow from how the product works.

Grounded guidance can still be fed a wrong value the model extracted from the chart, and the company describes no check on extracted values before they fill a calculator. The discharge module benchmarks length of stay and flags patients due for discharge with no readmission counterweight published, and no monitoring for premature discharge is described.

Mechanisms exist and nothing stands behind the output. The pathway feature presents supporting evidence from the patient's chart next to its answer, and ClinicalKey AI, which Elsevier plans to combine with the product, is described as tracing every response to its source. No error rate, omission measure, warranty, indemnity or remediation commitment is published.

The site footer states that certain described features are illustrative of potential capabilities and not in the product, without naming them, so it is not clear whether the evidence display ships in the version on offer.

Named record systems with production sites, and no write path described. Wellsheet names Oracle Cerner, Epic and athenahealth, says it is available on the Oracle Cloud Marketplace, Epic Showroom and athenahealth Marketplace, and names deployments at RWJBarnabas Health (seven hospitals) and AllianceChicago on athenaOne; the interface embeds in the EHR through APIs.

Generated documentation is described as a starting point for the clinician, but how it reaches the record is not described, the Epic and Cerner implementation guides are behind a form, and the athenahealth listing page is script rendered.

A single hosted option with the location implied. The product pages describe a fully cloud, API-based, web-delivered architecture with no interfaces to build, and the privacy notice mentions cloud hosting providers without naming them. No provider, region or residency commitment is published. Integration into Elsevier may move hosting and subprocessors, and there is no word on a planned migration or on the notice customers would receive.

No price is published, and the sales channel is discoverable. The company sells to health systems directly and says it is available on the Oracle Cloud Marketplace, Epic Showroom and athenahealth Marketplace. Nothing states the unit of charge, tiers or implementation fees, and the footer disclaimer means the scope being priced must be confirmed.

Inpatient care teams first, with ambulatory use named. The product supports multidisciplinary rounding across physicians, nursing and case management, inpatient documentation and discharge planning, and outpatient clinics, including AllianceChicago's community health centers on athenaOne. Named deployments stand behind the inpatient work, but no age, specialty or population exclusions are published.