A Startup Wants to Fix Health Insurance With an Engine That Learns From Denials
Insurf emerged from stealth in mid-August with an audacious pitch: software that could save employers $5,000 per worker annually by tackling two of the messiest problems in American healthcare at once.
The San Francisco startup, which recently completed Y Combinator's accelerator program, is betting that linking plan-selection tools with denial-fighting software creates something more valuable than either piece alone. One product, Surely, ranks individual marketplace health plans by projected 12-month cost. The other, Inveto, helps clinics navigate prior authorizations and appeals. The twist is that Inveto's real-world outcomes feed back into Surely's recommendations, letting employers and their workers see not just which plan has the lowest premium but which one denies the fewest claims.
Whether that feedback loop constitutes a genuine moat remains to be seen, but the timing is deliberate. Insurers are under growing scrutiny for authorization hurdles and claim denials, while a small but expanding slice of employers are exploring individual-coverage health reimbursement arrangements as an alternative to traditional group plans.
Surely ingests premiums, deductibles, drug formularies, and provider networks for individual-market plans across 16 states. It runs a deterministic simulation for each household, accounting for prescription needs and anticipated care, then surfaces the plan with the lowest total annual spend. According to the company, every input comes from government filings. The startup offers an example: one simulated household faces an $8,052 swing in a bad year by choosing the plan with the cheapest monthly premium but the wrong deductible and network.
Inveto, meanwhile, drafts prior-authorization requests with source citations, coordinates physician attestation, and tracks deadlines. Staff at participating clinics retain control over submission; Insurf does not file on their behalf. The product is live with clinics in Texas, Georgia, and California, though the company declined to name customers publicly. Insurf offers an accountability contract with fee credits if it misses a deadline and separates payer approval from actual reimbursement, a distinction that matters when green lights don't always translate into payments.
Co-founder and CTO Bryan Chung described the linkage between the two products as a "coverage-decision graph" in a launch post. Each de-identified outcome from Inveto becomes a data point for Surely. "Everything Inveto learns, Surely knows," he wrote, a feedback loop that the company says sets it apart from standalone decision-support tools or appeal vendors operating in isolation.
"American health insurance is f---ed: chosen by guesswork, enforced by denial," Chung wrote. "We built the engine that fixes both halves."
The company targets employers considering individual-coverage HRAs, a structure the Department of Labor projected in 2019 could eventually reach 800,000 employers and 11 million workers. ICHRA adoption remains modest compared to traditional group coverage, but Insurf argues that employees navigating these arrangements need better tools for both enrollment and disputes.
The startup's projected $5,000 annual saving per employee, according to Insurf, benchmarks against the Kaiser Family Foundation's employer health benefits survey, which found average premiums of $9,325 for single coverage and $26,993 for family plans. Insurf said the estimate assumes better plan selection and fewer surprise denials, though the methodology behind the figure has not been disclosed.
Co-founders Chang Lu and Chung met during a recent Y Combinator batch. Lu attended Brown University's combined BS/MD program and conducted zero-shot modeling research at Emory University's Winship Cancer Institute. Chung's profile lists international rankings in physics and computer science, two IEEE publications, and a top-tier competitive programming credential. The startup incorporated in Delaware and registered in California in mid-June. As of late August, the team numbered three people.
The launch arrives as insurers contend with renewed attention on prior authorizations. Medicare Advantage plans made nearly 53 million prior-authorization determinations in 2024, denying 7.7 percent, according to KFF. Only a small fraction of denials were appealed, but more than 80 percent of those appeals succeeded fully or partially. In the ACA marketplace, insurers denied roughly one in five in-network claims; fewer than 1 percent were appealed, and about a third of appeals overturned the denial.
An American Medical Association survey found physicians and staff spending an average of 13 hours weekly on prior authorizations, handling about 40 requests per week, with two in five practices assigning staff exclusively to the work.
Insurf enters a market with established competitors. Jellyvision's ALEX tool has guided benefit elections for years and recently extended decision support beyond open enrollment. Nayya offers plan-selection software and has acquired companies to bolster its AI credentials. Picwell provides predictive APIs for employers and benefits administrators. Take Command administers ICHRA and QSEHRA plans and claims a client base in the thousands.
Insurf said it does not operate as a licensed insurance agency, does not sell insurance, and receives no carrier commissions. Enrollment applications flow through an individually licensed producer. The distinction matters for both regulatory exposure and revenue model; the company's fees come from employers or clinics, not insurers.
On the compliance front, Insurf has not yet issued a SOC 2 report but says it is working toward readiness. Production handling of protected health information is currently disabled until business-associate agreements, vendor contracts, and technical safeguards are complete. The startup lists AWS, Vercel, Neon, and OpenAI as vendors in the PHI data path.
The company's privacy notice covers its public site, Inveto, Surely, and a Chrome extension called "Insurf PA Copilot." No public Chrome Web Store listing for the extension was located; it may be distributed privately during the pilot phase.
Surely is available to select partners with an illustrative demonstration, while Inveto remains in a private pilot. The company's status page listed the public site as operational, Inveto as in "private pilot readiness" with production PHI disabled, and Surely as running public vertical slice and private evaluation workflows.
"Surely computes the lowest-cost plan for anyone. Inveto fights denials for real clinics across Texas, Georgia, and California, and each contract-authorized, de-identified outcome becomes coverage data nobody else has," Chung wrote. The company is eyeing this year's open enrollment period, which begins in November.
Whether linking plan selection to denial outcomes proves defensible as a competitive advantage will depend on scale, clinic adoption, and how quickly larger players notice. For now, Insurf is building in a regulatory gray area where most tools treat enrollment and claims as separate problems. The bet is that solving both, and connecting the data, matters more than the founders expected it would.
