Forus

The models do the core work, and a network and workflow platform sit around them. Forus generates and submits prior authorizations and drafts appeals from the patient's record, and it also routes prescriptions to pharmacies, verifies benefits, handles enrollment forms and specialty pharmacy calls, tracks renewals, finds affordability programs and keeps patients updated, with a dashboard that follows each prescription.

The company describes itself as building an AI-powered network connecting doctors, pharmacies, payers and biopharma. Take the models away and the routing, tracking, patient messaging and network relationships remain, so the models are the engine of the authorization and appeal modules rather than the whole of what is offered.

The system acts rather than recommends, and the point at which a person approves a submission is not published. Forus says a clinician writes the prescription in the record as normal and Forus automatically generates and submits the prior authorization form, checks status with the plan, and routes the prescription to a pharmacy.

When a request is denied the practice chooses the next step, such as starting an appeal that Forus drafts, and staff can follow every prescription in real time and step in. The company's engineering blog describes a language model judge that checks each form answer against the patient's record and internal guidelines before submission. The company does not say whether anyone at the practice reviews a form before it goes to the payer, what the judge does with an answer it cannot support, or how often it intervenes.

Technique is described, and provenance is not. Forus's engineering blog describes a language model judge that verifies form answers against the patient's record and internal guidelines, and a retrieval-augmented generator for appeal letters, which separates writing an answer from checking it.

No underlying model provider, model version, hosting arrangement or accuracy figure for either component is published, and nothing says whether the judge and the generator share a model family, in which case they may share blind spots.

The architecture is outlined, and no party in it is named. Forus describes a generator that completes authorization forms and appeals from patient records and a separate judge that checks them, and its privacy policy says service providers and partners may access personal information in operating the service and that anonymized data and aggregated insights may be sold to third parties.

No model provider, hosting provider, fax or e-prescribing network, or subprocessor list is published, nor the retention and training terms that apply to them. Full clinical records pass through the generator and the judge, and prescriptions pass on to pharmacies and plans.

Named practitioners report throughput and turnaround gains, and none is measured against a baseline. A biologics coordinator at Digestive Health Specialists reports working 15 to 30 authorizations a day, up from five; a practice manager reports phone time falling from 60 to 70 percent of the day to about 15 percent; clinicians at Optima Dermatology, Goodman Dermatology and Unio Specialty Care describe fewer rejections and faster approvals; and case studies cover an allergy and rheumatology practice and a group purchasing organization. None states a period, a denominator or a first-pass approval rate, and approval gains are not separated from staff throughput.

General assurances are published, and the questions an authorization product raises are not answered. Forus says its platform is HIPAA compliant and that all protected health information is covered by a business associate agreement.

Its privacy policy, updated May 2026, says it may create anonymized and aggregated information and 'may sell anonymized data and aggregated insights to customers and other third parties', and keeps personal information as long as needed for its purposes, including pharmacy regulations. Nothing says how long extracted patient records are kept once an authorization is decided, whether records train or tune the generator or the judge, or how de-identification is done before data is sold.

The platform reads full clinical records to complete forms, so that surface is broad. Safety engineering, meaning guardrails, hallucination mitigation and how a safety event is handled, is described only through the verification judge.

Business associate coverage is stated plainly, and the agreement is not published. Forus's FAQ says the platform is fully HIPAA compliant and that all protected health information is protected under a business associate agreement, with no tier or paid upgrade attached, which matters for a platform that is free to providers.

The agreement itself, its subcontractor terms and breach notice period are not published, and the privacy policy that is published covers the website and describes selling anonymized data and aggregated insights.

A SOC 2 Type 2 attestation is named as the company's own, with its reports held in a trust portal. Forus's FAQ says the platform is SOC 2 Type 2 certified with enterprise-grade security, alongside its HIPAA and business associate statements, and the site's Security link leads to a trust portal at security.forus.com. The public pages do not state the report period or whether documents open without a request. No HITRUST or ISO 27001 certification is named.

No device pathway applies. The platform automates administrative and reimbursement work and does not make or inform a diagnostic or treatment decision, so it sits outside software as a medical device. The rules that do apply are on the payer side: the CMS interoperability and prior authorization rule reshapes electronic authorization requirements over the coming years, and each plan sets the criteria a submission must meet. Forus publishes no statement on how it will meet the CMS timeline.

One output check is described, and no governance framework is. Forus's engineering blog describes a language model judge that verifies generated authorization answers against the patient's record, which tests model output in production; no governance framework, model monitoring program or bias evaluation is published, and no result from the judge is.

The concern is specific to this product: a system that drafts clinically grounded authorizations and appeals could systematically do worse for patients whose records are thinner, and those are disproportionately people who have moved between health systems, gone through uninsured periods, or received care where documentation is weakest. An unmeasured gradient would fall hardest on the patients a medication access platform exists to help.

A verification mechanism checks generated answers before submission, and nothing stands behind a wrong one that gets through. Forus's engineering blog describes a language model judge that checks each authorization form answer against the patient's record and internal guidelines, so a separate component tests the generator's claims; practices also see every prescription's status and choose whether to appeal a denial.

No accuracy figure for the judge, error rate, warranty, indemnity or remediation commitment is published, and the website terms disclaim warranties for the website. The characteristic harm is an authorization submitted with a clinical assertion the record does not support, under the prescriber's name, which can trigger a payer audit or a denial that delays therapy.

Integration is claimed broadly, with no record system named. Forus says it integrates with all major or cloud-based record systems and picks up the prescription at the point of electronic prescribing; practices whose record system does not integrate can upload records when asked, and practices that do not e-prescribe are not yet supported. No record system, e-prescribing network, marketplace listing or interface method is named, and what, if anything, Forus writes back into the record is not described.

No public disclosure of hosting arrangement, cloud provider, data residency, or whether a private tenancy option exists. The platform is evidently cloud delivered given the lightweight setup claim, but nothing beyond that is stated.

That silence covers a product that sits inside the EHR at the point of prescribing.

The price to providers and patients is published and it is zero, and the other side of the exchange is published only in outline. Forus says it is free for providers and patients, with no trial qualifier attached. The company describes itself as a network connecting doctors, pharmacies, payers and biopharma, and its privacy policy says it may sell anonymized data and aggregated insights to customers and other third parties. What data derived from a practice's prescriptions and records is sold, to whom, and whether any party that pays Forus influences pharmacy or program choice is not published.

Coverage is stated as any drug, any insurance and any pharmacy in all 50 states, across medical and pharmacy benefits, specialty pharmacy and buy-and-bill, for practices from solo independents to health systems. Named customers and case studies concentrate in specialties where authorizations run high: gastroenterology, dermatology, allergy and rheumatology, with a group purchasing organization among them.

Validation is testimony rather than measurement, and only practices that prescribe electronically are supported. Forus works on the requesting side of prior authorization, where it generates and submits authorization forms from the prescription and record, checks status with the plan, drafts appeals, tracks renewals and verifies benefits.