The 47-person AI company training machines to read the insurance industry's least-loved paperwork - medical files, financial records, and reinsurance treaties - and hand the answers back, sourced.
Walk into the underwriting floor of any large life insurer and you'll find the same scene: a queue of PDFs, a stack of faxed lab reports, a doctor's note written in a hand that hasn't been legal tender since 1994. Somewhere, a human is being paid to retype it.
Friendly's pitch is that this scene doesn't need to exist. The company - founded by Natasha Alexeeva and headquartered in the Financial District - has spent the better part of six years building one thing: an AI infrastructure layer that speaks insurance. Not a chatbot. Not a wrapper. A pipeline that takes whatever the carrier received - medical records in three languages, financial statements scanned at 200 DPI, treaty amendments with redlines from 2008 - and returns structured, sourced, auditable answers.
It is a small bet on a large premise: that the bottleneck in insurance is not capital, not actuarial science, not even regulation. It is paper. And paper, unlike most things, is something machines have finally learned to read.
Most generic large language models, asked to summarize a 600-page claim file, will write something fluent and partially invented. That's a problem when a misread becomes a denied claim. Friendly's wager is that the industry will pay for the boring version: every extraction tagged to a source page, every confidence score visible, every decision reproducible by the compliance team. The chat assistant sits on top, but the boring receipts are the product.
Friendly's product line is narrow and stubborn. It does the same trick - turning unstructured documents into structured insight - across the three places insurers leak the most time.
Digitizes medical and financial records, extracts risk signals, flags misrepresentation. Built for STP - straight-through processing - of new business.
Handles workers' compensation, disability, and critical illness claims. Generates chronological case timelines and narrative summaries.
Treaty analysis, amendment tracking, and term comparison across decades of agreements. Reinsurers' last spreadsheet, finally indexed.
Friendly is staffed unusually for an AI startup - half the senior team has spent careers inside the companies it now sells to.
15+ years in machine learning, data analytics and software architecture. Started Friendly to put research-grade AI behind problems that don't make headlines.
25 years across AI and software engineering. Leads the model and pipeline stack.
Former CEO of PartnerRE and Scor Global Life RE - the kind of reinsurance pedigree that opens doors.
Veteran of Munich Re, Swiss Re and GE Financial. Translates assessor instincts into product spec.
A look at what's moved through Friendly's press room in 2025.
The company expanded its claims division, hiring assessors and product owners to back the auto-adjudication platform.
A reinsurance veteran joins to drive adoption of Omniscient, the treaty management platform.
Xceedance backs a follow-on round; total funding sits at $1.29M across the company's life.
Six years on, the stack of faxed lab reports is still there. The doctor's handwriting hasn't improved. What has changed is who reads it first.
At a handful of carriers and reinsurers - Swiss Re, PPS, GroupHealth, MunichRE among them - the first pass is now a Friendly pipeline. Pages go in. Structured fields, source citations and a confidence score come out. The underwriter still signs the file. The assessor still calls the doctor. The reinsurer still negotiates the treaty. But the queue is shorter, and the receipts are better.
Friendly hasn't deleted the underwriter's inbox. That was never the bet. The bet was that an AI built specifically for the language of insurance - patient and audited and unimpressed by its own fluency - could quietly empty most of it. So far, the inbox is shrinking.