Sila Health raises $27M to automate healthcare prior authorization with AI — Pavle Asparouhov and Delian appear together

Summary

Sila Health raised $27 million in a $5 million seed and $22 million Series A led by Bain Capital to automate healthcare prior authorization with AI. Pavle Asparouhov, a former early employee at Ramp where he worked on bill pay, co-founded the company alongside Jeff Morelli, the CEO, who has a background in healthcare software and previously worked with Asparouhov building clinical workflow tools for autism therapists.

Prior Authorization as a Friction Point

Over the past decade, healthcare shifted from post-care billing to requiring insurer sign-off before care is delivered. The change was designed to prevent surprise bills and unnecessary procedures but created significant administrative drag. A single clinician may deal with ten different insurers, each running separate approval processes. The paperwork burden pulls clinical staff away from patients, and delays in authorization can directly delay care, particularly for vulnerable populations like children in autism therapy.

Sila's product ingests patient insurance data, determines what clinical documentation is needed, validates it against insurer requirements, and routes clean submissions. This removes the need for staff to manage the process manually. Asparouhov frames it as a data-movement problem rather than a systemic healthcare flaw. Insurers have reasonable processes, but fragmentation across carriers means the coordination cost falls entirely on providers.

Operating Scale

Sila operates in 45-plus states, works with hundreds of insurers, and has processed authorizations for tens of thousands of patients. Asparouhov deliberately delayed the company's public launch until it could announce those numbers. In an AI hype cycle, tangible operating metrics carry more weight than pre-revenue announcements.

Why Healthcare Over Fintech

Asparouhov left Ramp because finance teams are largely satisfied with their technology stack. Healthcare data is fundamentally unstructured and textual, so the previous wave of SaaS tools never delivered a 10x improvement for providers already working in person with physical notes. LLMs change that equation, making it possible to extract and route clinical information in a way that genuinely transforms the workflow rather than digitizing an existing one.

Series A Use of Proceeds

The $22 million will fund engineering and sales hiring, expansion into additional specialties, and deeper product functionality. Sila currently handles the clerical side of prior authorization. The next phase involves using the company's growing familiarity with insurer-specific guidelines to help providers submit documentation that preemptively addresses common rejection triggers.

AI SDR Critique

A private equity investor and former head of AI at a billion-dollar PE fund, posting on X as "Carried No Interest," argues that the AI SDR category is structurally broken, not just for 11x but for the concept broadly.

Cold email is an alpha game requiring constant adaptation to platform quirks. Abstracting that away through a managed product erodes the edge. Five compounding problems emerge: dependence on a third-party contact database that compresses gross margins; high probability of embarrassing both vendor and customer through poorly targeted outreach; no testing environment to catch mistakes before they cause reputational damage; complex onboarding requiring a full customer success operation; and churn rates that don't resemble enterprise software benchmarks.

Cursor's unit economics offer a stark contrast. It is self-serve, requires zero marketing spend, has low ACV, low CAC, and high stickiness. That profile reaches $100 million ARR in two years. AI SDR companies sit in a worse position on every dimension: no self-serve, high CAC, compressed margins, and bad retention.

The investor sees value in LLMs automating market mapping and persona identification, but that is a feature of ZoomInfo or Apollo rather than a standalone venture-backable company. A human-in-the-loop co-pilot model probably makes more sense than a fully autonomous outbound agent. He expects 50 companies are already building that.

Cursor's Durability

On whether fast-growing AI tools like Cursor can build durable moats, the investor is skeptical about an IPO path. The Slack-Teams dynamic worries him: a best-in-class product that loses ground the moment a large distributor bundles a good-enough competitor. Cursor is more likely to be acquired at a significant valuation than to sustain a standalone public company, given churn risk and the threat from Microsoft or other large tech platforms with existing distribution. He acknowledges he could be wrong.