Ascertain wants to get healthcare out of the hold queue

A hospital discharge has two kinds of readiness. There is the medical kind, decided by people trained to care for patients. Then there is the administrative kind: the records, the destination, the insurer’s requirements, the authorization. Ascertain has built its business in the space between them. It is a peculiar place for an AI company to seek its fortune. The action happens in portals, forms, and follow-ups. The patient is waiting somewhere else.

That separation explains the appeal. A clinical decision can be excellent while the machinery around it is inefficient. Ascertain sells to the organizations running that machinery: health systems, health plans, and provider groups. For a patient, its useful work would often be invisible. For the employee preparing another submission, invisibility would be quite welcome.

A company factory finds a specific job

Ascertain’s 2022 launch described a company-creation platform, a collaboration between Northwell and Aegis Ventures. The ambition included maternal health, disease detection, and health equity. Its ingredients were unusually practical: a venture studio, clinical expertise, data, and a health system where ideas could meet actual work.

The early venture-building effort helped launch Optain, a retinal-imaging company, in 2023. Ascertain then appointed Mark Michalski, a radiologist whose experience included Amazon, Mass General Brigham’s clinical data science center, Butterfly Network, and Hyperfine. He had worked on both sides of the argument: what technology can do, and what healthcare organizations can absorb.

In a 2025 interview, Michalski said early AI-agent demonstrations at Amazon suggested a use for healthcare’s routine administrative tasks. He brought that thesis to Ascertain and retooled the company around it. The shift gave a broad venture-building ambition a specific customer and a specific job. It also supplied an unusually sober test: could the work get done with less effort?

The portal is where the work lives

Ascertain’s current platform handles tasks including eligibility verification, prior authorizations, documentation, referrals, and care coordination. The Unified Payer Portal, or UPP, is a particularly revealing product. In outpatient oncology, it automates data entry, documentation submission, and navigation through payer portals before a visit. Those are small actions until somebody must repeat them all day.

Its market position rests on handling a sequence rather than offering an isolated answer. An organization can already hire staff, outsource work, or install conventional automation. Ascertain argues that its agents can interpret varied inputs and adapt to the messy workflows that make fixed scripts brittle. That is the company’s competitive claim; the reported deployments give it something concrete to discuss.

In his Leerink interview, Michalski described a framework of specialized agents working toward a shared goal. Breaking work into parts is supposed to make corrections easier to locate. An incorrect rule or diagnosis code becomes a problem with an identifiable place in the process, rather than a mysterious answer delivered by one enormous machine.

“It makes it very intuitive; it makes it resilient, it makes it auditable, it makes it correctable.”Mark Michalski, speaking to Leerink Partners

Count the clicks. Keep the units.

At Northwell, Ascertain worked with the centralized transitions-of-care administration team on processing authorizations. Its reported pilot results were an 80% reduction in user clicks and a 67% reduction in authorization submission time. These are appealingly mundane measures. Someone had to count what the old process demanded before declaring the new one useful.

The Oncology Institute supplied the next useful test. A signed statement of work became an early live deployment in eight weeks; it went live in September 2025. The partners reported more than an 80% cut in office-visit authorization submission time at pilot sites. Their November announcement said the system was processing authorizations across more than 100 clinics and affiliate locations.

The same announcement estimated up to $2 million in operating expense savings in 2026 as the work expanded. The wording matters. An estimate is an invitation to measure again. Faster submission also leaves another clock running: the payer still has to decide. An administrator’s saved minutes and a patient’s treatment date belong in different columns.

Capital buys expansion; customers buy execution

Ascertain’s April 2025 $10 million Series A was led by Deerfield Management, with Northwell participating. That financing should be read separately from the original $100 million commitment associated with building multiple healthcare AI companies. Combining the figures produces an impressive number and a less useful understanding of the business.

The commercial offering combines enterprise software with technology-enabled operational services. The buyer purchases help carrying out work, with healthcare expertise around deployment and support. In January 2026, CVAUSA announced a partnership covering a network described as more than 100 clinic locations across eight states. That is a new setting for the approach, rather than proof that every site produces identical results.

The part another organization can copy

The transferable method is modest: pick one workflow, establish a baseline, deploy a bounded pilot, and expand after measuring it. Ascertain’s own writing about cautious buyers treats diligence as a rational response to earlier technology disappointments. The company’s public philosophy favors co-development and evidence over a grand presentation. A buyer can borrow that discipline without borrowing the software: choose an observable task, agree what success means, and ask who will handle the cases that automation cannot finish. A small experiment is useful precisely because its result can be checked. That makes the operator’s judgment part of the product.

The approach needs access to the right systems, usable records, an owner for the workflow, and people available when the routine becomes an exception. Ascertain includes human review and escalation in its published operating model. Automation cannot grant an insurer’s approval or make incomplete records complete by wishing. The useful question is whether it removes enough work to justify what the organization puts into it.

Ascertain’s most interesting proposition is therefore easy to picture: a trained person spends less time shepherding a form through a portal. The machinery becomes less demanding. Somewhere beyond the screen, there is a patient who did not come to healthcare for the paperwork.