Innobot Health

Much of what Innobot sells would run without a model. RIS moves eligibility and claim status transactions through a clearinghouse, ICORE builds and IHELM runs robotic process automations that log in to payer portals and EHRs the way a person would, and IFLOW mines desktop activity to find work worth automating; next steps are chosen by an algorithm that reads the latest claim event.

The learned parts sit in the agents that converse. IVERA holds phone calls with payer phone systems and turns them into transcripts with a confidence score, ICHAT talks to payer portal chatbots, and IKNOW answers questions about a customer's claims in plain language. Those agents extend what the bots can reach, and the revenue cycle automation around them stands on its own.

Each task carries its own route to a person. Work moves through APIs, portal bots, chat agents and voice calls, and anything every layer fails on falls into an exception queue for staff to resolve, with owners, due dates and dispositions tracked. Voice call results come back with a transcript, a confidence score and recommended next steps. On prior authorization, agents handle 85 to 95 percent of routine requests and route exceptions to the customer's team with the documentation already gathered, and IHELM logs every run.

The platform says completed tasks need no human involvement, yet the confidence level below which a result goes to a person rather than being written back is not stated, and neither is what happens when a wrong status or payment is posted.

Agents are named by job, IVERA for payer phone lines, ICHAT for portal chatbots and IKNOW for questions about a customer's own revenue cycle data, and described as applying contextual reasoning and scoring their own confidence. Which speech recognition, language or reasoning models sit underneath, whether they are built in house or licensed, and how versions change are not stated.

The automation layer is clearer than the model layer: ICORE generates coded, tested robotic process automations from a process definition document, and IHELM schedules, runs and logs them, so a buyer can follow how the scripts are built and run without learning what powers the conversations.

One outside party is named: the Stedi clearinghouse carries RIS's eligibility and claim status transactions. Beyond that, the trust page says every vendor with access to systems or data works under formal agreements covering security and breach notification, and every contractor with system access signs a business associate agreement.

No subprocessor list exists, and the model layer behind the voice, chat and knowledge agents is unnamed, as are the hosting provider and any telephony service carrying IVERA's calls to payers. The privacy policy names analytics tools such as Google Analytics and Microsoft Clarity, and those apply to the website only.

Results are reported as production figures from unnamed clients: $1.29 million of financial benefit for a national billing organization, $1.16 million for a multi state dermatology group, 4,500 claim status checks a month for a surgery center billing company, and 2,000 to 4,370 hours saved for home health, dental and dermatology groups.

The homepage adds a 667 percent return at month eight of an enterprise deployment with 75 full time staff freed, 97.3 percent less scheduling time and 99.8 percent automated reconciliation, and the founder's biography cites more than $1.3 billion in added client revenue. None comes with a client name, method, baseline or period, and the dashboard figures on the platform page are labeled example data rather than client results.

Protection of patient data is described as security controls rather than AI terms: AES 256 encryption at rest, TLS 1.2 or higher in transit for claims, ERA files and integration payloads, role based access with multi factor authentication, and audit logging.

What the agents do with the data goes unaddressed. Nothing says whether payer call transcripts, portal chats or claims data train any model, how long call recordings and transcripts are kept, or whether data reaches an outside language model. The retention schedule in the privacy policy covers website data, leaving client PHI to each business associate agreement. No safety engineering is described for the voice and chat agents that speak with payers on a client's behalf.

Innobot handles protected health information only as a business associate of its healthcare customers, under a signed agreement and inside contracted service environments, and says so in plain terms. An agreement is included with every client contract at no extra cost, and the standard version sits in a HIPAA compliance package, alongside a risk assessment summary and workforce training attestations, available on request.

The trust page adds that every contractor and third party with system access signs a business associate agreement before access is granted, with PHI access controls, audit logging, breach notification procedures and annual risk assessments in place. The agreement itself cannot be read before asking, so its scope and subcontractor terms are known only from that summary.

SOC 2 Type II is named for the company's own platform, covering the security, availability and confidentiality criteria, audited by an independent CPA firm and renewed annually, with the full report available to prospects and clients under NDA through a request on the trust page. The audit firm and the report period are not given.

Controls are described in detail: vulnerability scanning and regular penetration tests, annual incident response tabletop exercises, change management with peer review, least privilege access reviews, endpoint encryption, security training and cybersecurity insurance. Access to the report runs through Innobot's own approval rather than a self serve portal.

Administrative work keeps this product clear of device questions. Innobot automates eligibility and claim status checks, assembles and submits authorization packets on the provider's side, scrubs and resubmits claims, posts payments, and flags inpatient stays for utilization review as they near the 24 and 72 hour marks.

Agents gather clinical documentation to attach to authorization requests without judging medical necessity, which stays with the payer, and the hospital pages say the clinical decision stays where it belongs. No device claim is made and no regulatory position is stated beyond that, the usual footing for revenue cycle automation that never touches diagnosis or treatment.

Nothing describes how the agents' behavior is tested or governed. The SOC 2 controls cover change management, peer review and testing for production deployments, which governs software releases rather than what a model says or captures.

There are no accuracy evaluations for IVERA's call transcripts or IKNOW's answers, no account of how confidence scores are calibrated, and no review of how the voice and chat agents perform from one payer's phone tree or portal to the next. The company's principle of telling clients when a process is not ready for automation is an engagement promise rather than an evaluation of the models.

Mechanisms exist to catch a bad result: every agent run is logged in IHELM, voice calls leave a transcript and confidence score attached to the task, and exceptions go to staff with their documentation.

Nothing stands behind the output itself. Captured statuses, submitted authorizations and posted payments come without a stated error rate, and the guaranteed return within three to five weeks is a commitment about financial results that names no remedy if posted data turns out wrong. Maintaining automations when payer portals change is included in the service, which keeps the bots running rather than vouching for the accuracy of their work.

Integration happens at the screen. Innobot's agents log in to Epic, Cerner, Meditech, NextGen, eClinicalWorks, athenahealth, AdvancedMD, ModMed, Net Health, WebPT and other systems over remote desktop or VPN, run reports, update patient accounts and enter notes as staff would, so no API project is needed. Data can also arrive through an API connection or a scheduled export, and eligibility and claim status transactions run through the Stedi clearinghouse.

The record systems are named, with reading and writing at the user interface level, yet no marketplace listing, partner certification or named site confirms any of them in production, and screen level automation depends on interfaces that change with each release.

IRIS runs as a hosted platform from contracted service environments, with agents reaching each customer's systems over remote desktop or VPN. The company is based in the United States and runs its website there, but no hosting provider, region or tenancy model is named for client data, and where call recordings and transcripts are stored is not stated.

Encryption covers data at rest and in transit, a configuration management system tracks production infrastructure, and capacity reviews run regularly. The credentials the bots use to sign in to customer EHRs and payer portals are part of every deployment, and how they are stored, scoped and rotated is not described.

Pricing follows transactions rather than seats: RIS, the core licensed platform, is billed per transaction, and an illustrative dashboard shows $0.14 per transaction, marked as example data rather than a price. No actual rate, minimum or tier is given.

Engagements start with a single process delivered in three to five weeks, with a guaranteed return and no long commitment, then grow into a roadmap across other workflows, and keeping automations working when payer portals change is part of the service. A buyer can learn the unit of charge and the shape of a first engagement before a demo, but not what a transaction or a build costs.

Three buyer types are named: hospitals and health systems, provider groups and practices, and RCM and billing companies. Case studies span a national billing organization, dermatology groups, a surgery center billing company, a home health and hospice provider and a multi location dental group, and the inpatient material covers utilization review flags at the 24 and 72 hour marks, bed occupancy and nursing ratios.

Workflows run across prior authorization, eligibility, claim scrubbing, denials, payment posting, charge capture, scheduling and reporting, with automations already built for Availity, Optum, PaySpan, NaviNet, the Medicare administrative contractors and major commercial payers. The case studies are anonymous, so support for any one setting cannot be traced to a named organization.