Waystar

AI is a layer across a revenue cycle platform whose value stood before it. Waystar's AltitudeAI spans predictive scoring (which denials to work first, charity care screening, propensity to pay), generative drafting (appeal packages, clinical rationale for coding denials, medical-necessity summaries for authorizations) and, since 2026, agentic features that act within workflows, all across a claims, eligibility, remittance and payments platform that clears transactions for about one million providers. A provider organization buys the platform; the AI accelerates work inside it. With the models removed, the clearinghouse, payments and claims management remain.

Autonomy is claimed and oversight appears as a phrase. The home page describes agentic AI that 'continuously acts within workflows, executes defined tasks, and learns from outcomes with minimal intervention' and generative AI producing 'autonomous content'; the Innovation Lab describes appeal packages generated 'while staying in control of the process', a billing concierge that 'seamlessly escalates to human staff when necessary', autonomous Medicaid eligibility monitoring that routes patients, and agentic support that helps CDI nurses judge whether a working DRG fits the clinical picture.

No review step before an appeal or authorization summary reaches a payer, threshold, escalation rule or error rate is published. The home page says AltitudeAI is 'purpose-built to ... eliminate errors', a statement of aim about the revenue cycle rather than a claim that its own output is always right, and recorded as such.

The approach and the inputs are named; the models and how well they work are not. The company's filing describes a rules engine with machine learning layered over the transaction flow, trained on the billions of transactions the platform carries, and names the applications: predicting whether an appeal will succeed, finding missing charges, aligning claim status with expected remittance, prioritizing work queues by expected value, predicting charity care eligibility and modeling patient payment behavior; it names the data elements too, including demographics, geography, diagnosis and prognosis, care provider and counterparty details.

The generative and agentic features (appeals, clinical rationale, authorization summaries, billing concierge, CDI support) are described by function. No model, generative provider, version, accuracy figure or update practice is published.

Hosting providers are named; the model layer is not. The company's filing names co-located data centers plus Microsoft Azure, Amazon Web Services and Google Cloud as its infrastructure, and describes predictive models trained in house on its own transaction data.

The generative and agentic features (appeal drafting, clinical rationale, authorization summaries, the billing concierge) are not attributed to any model or provider, and no subprocessor list is published, so a buyer cannot tell which outside services receive claims and clinical documentation. Its published business associate agreement requirements include equivalent agreements with subcontractors.

Named deployments with outcome figures, mostly for the platform rather than its AI. Customer stories on the home page credit Piedmont with more than $10 million in payment lift, Mount Sinai with a 300 percent increase in back-office automation, Renown Health with a 50 percent reduction in patient AR days, Proliance Surgeons with twice the patient payments, Cincinnati Children's with a 50 percent decrease in clearinghouse costs and SCP Health with 65 percent fewer days to pay, and University Hospitals and Advocate Health are named for EHR integration and AI financial clearance.

For the AI features the company reports appeal packages created three times faster (about 16 minutes saved each), more than 40 percent higher appeal overturn rates, 85 percent auto-approval on prior authorization submissions and a 40 percent cut in manual correction work in early agentic deployments, without method or named customer. A 2023 third-party survey reports 94 percent satisfaction with automation and EHR integrations. Nothing is independently evaluated.

The position is stated, and it runs toward pooling. The company's filing says insights generated from every transaction are embedded in platform updates deployed across its entire client base, so one organization's transaction data improves a product its competitors use, and it names demographics, geography, diagnosis and prognosis among model inputs.

It publishes what its business associate agreements must contain, including safeguards, subcontractor agreements and incident reporting. The website privacy policy covers site visitors, keeping web analytics data 24 months and no longer than five years, and says support staff are trained and bound by confidentiality.

No retention schedule for transaction or clinical data, de-identification method for pooled insights, or opt-out from pooling is published, and patients subject to charity care and collection models have no stated visibility.

Both HIPAA roles are stated and the agreement's contents are described; the agreement itself cannot be read before contracting. The company's annual report says Waystar is a covered entity when providing clearinghouse services and a business associate, or subcontractor to one, when providing software to providers and revenue cycle companies, and describes what its business associate agreements must contain: written assurances on use and disclosure, administrative, physical and technical safeguards, equivalent agreements with agents and subcontractors, incident reporting and assistance with the customer's own HIPAA duties.

It also addresses breach notification, 42 CFR Part 2, reproductive health privacy and online tracking guidance. The website privacy policy includes a HIPAA section. The agreement text and any tier limits are not published.

Recognized certifications are named in the company's own filings, and none is available without a request. Waystar's annual report on Form 10-K states HITRUST CSF certification (first received in 2021, with Coalfire named as assessor in the company's announcement), a SOC 2 Type II examination under SSAE 18, and PCI DSS Level 1 Service Provider status with validated point-to-point encryption for card payments, and names its external security auditors and vendors. The website footer links a responsible disclosure page. There is no trust center, and the HITRUST assessment level, the SOC 2 period and scope, and the certificates themselves are not published.

No device claim is made, and the regimes that govern the product are named rather than a device pathway. The company's filing describes it as a HIPAA covered entity when providing clearinghouse services and a business associate otherwise, operating under the HIPAA electronic transaction standards and PCI DSS, and maps its exposure under the Anti-Kickback Statute, False Claims Act, Stark Law and debt collection law; those are scope, not a device position.

Two AI functions sit near clinical judgment: Auth Clinical Summaries generates medical-necessity summaries from the record for authorization requests, and agentic CDI support surfaces clinical evidence for nurses assessing whether a working DRG reflects the patient's condition, with the clinical rationale generator drafting responses to clinical validation denials. Each is described as drawing on documentation already in the record for a reviewer.

No governance framework, bias evaluation, subgroup analysis or independent audit is published, and the company's own disclosures show why one is needed. Its filing names demographics and geography among model inputs, and names applications that make determinations about individual patients: predicting whether a patient qualifies for charity care to drive financial intervention before service, and modeling patient behavior to align collection costs; the Innovation Lab adds propensity-to-pay scoring and charity screening in financial clearance.

Geography in the United States is a close proxy for race and income, so a model that performs unevenly across groups would distribute both financial help and collection effort unevenly. Nothing published says which direction each intervention runs, whether accuracy varies across groups, or who reviews the models.

Staff remain in the loop on some AI outputs, and nothing stands behind the output. The Innovation Lab describes appeal packages generated while teams stay in control of the process, CDI agentic support that surfaces evidence for nurses to judge, and a billing concierge that escalates to staff; the filing names the models' applications and inputs, which lets an affected party ask how a determination was reached.

The published AI figures (higher appeal overturn rates, 85 percent auto-approval on authorization submissions, fewer manual corrections) measure throughput rather than correctness, and no accuracy, error rate, warranty, indemnity or remediation commitment is published. A wrong medical-necessity summary or a charity care score that misses an eligible patient carries no stated recourse.

Named integrations with the major record systems are documented, running in production at named health systems. The supported systems page lists the practice management and hospital systems Waystar integrates with, including Epic, Cerner Millennium, MEDITECH, eClinicalWorks, NextGen, Veradigm and more than a hundred others, and the company's filing reports integration with more than 500 record and practice management systems plus more than 200 channel partners; the Innovation Lab reports 1,200 payer connections for eligibility.

Data moves both ways: eligibility, claim status and remittance return into the provider's systems, and University Hospitals is named for its record-system integration. The Innovation Lab adds FHIR-based prior authorization connectivity for CMS-0057-F. A 2023 third-party survey reports 94 percent of clients satisfied with its EHR integrations.

Hosting and architecture are stated; regions are not. The company's filing says infrastructure spans co-located data centers and Microsoft Azure, Amazon Web Services and Google Cloud environments, on a single-instance, multi-tenant, event-driven platform built in house, with availability above 99.9 percent, and that all employees are located in the United States.

It also states that insights from every transaction are embedded in platform updates deployed across the client base, so the multi-tenant design is deliberate. Which workloads run where and in which regions is not published, and there is no subprocessor list.

The pricing basis and contract terms are published; rates are not, and there is no route in without a sales conversation. Waystar's annual report on Form 10-K says contracts generally combine a subscription fee with a volume-based component, sometimes only one; initial terms run two or three years and renew automatically for one-year terms with standard price escalators; billing is monthly; and more than 99 percent of revenue is recurring subscription or predictable volume.

It also discloses concentration (top ten clients 11.2 percent of 2024 revenue) and net revenue retention (110.1 percent). No price list, per-transaction rate or tier is published, and nothing says whether AltitudeAI features carry an incremental charge or sit inside the platform subscription. The site's route is a demo request.

The product line covers financial clearance and patient access: eligibility verification, coverage detection, prior authorization, registration quality, patient estimates, charity screening and Medicaid coverage monitoring, alongside claims, remittance and denial work.

Coverage is named and supported across settings: health systems and hospitals (Piedmont, Mount Sinai, Renown Health, University Hospitals, Advocate Health), children's hospitals (Cincinnati Children's), multi-specialty surgery groups (Proliance Surgeons), ambulatory surgery centers, billing services, laboratories, durable medical equipment, fertility, FQHCs, home health and hospice, and skilled nursing, each with its own page.

The company's filing reports more than 30,000 client organizations and about one million providers, with revenue split roughly 30 percent hospitals and health systems to 70 percent ambulatory and alternate sites. The AI features run across that platform; results specific to them are not broken out by setting.