Coral AI And Healthcare’s $450 Billion Paperwork Problem

Original Post : Source

New York-based Coral AI attacks healthcare’s most frustrating administrative work: the faxes, forms, portal logins, phone calls, and prior-authorization exchanges that delay patient intake and billing. Founded in 2024 by Ajay Shrihari, a robotics and AI researcher, and Aniket Mohanty, an expert in medical image processing, Coral’s model is trained on the industry’s vast trove of data to automate the data input and exchanges that clogs the system.

Coral reads long, messy referral packets, extracts relevant data, checks eligibility, reasons through clinical criteria, and drafts prior authorizations. Human staff review and submit. The company says its models reach accuracy levels in the high 90s and now handle hundreds of thousands of patient workflows monthly.

“We didn’t start with healthcare as an industry,” Shrihari said. “We started with the frustration of watching people wait weeks for care because of paperwork. Once we saw how much time was lost before a doctor ever saw a patient, the opportunity became obvious.”

Ashish Singh of Bain & Company, a senior advisor to Lightspeed Venture Partners on healthcare, was studying paperwork bottlenecks when he met Shrihari through Lightspeed’s network. He immediately saw how computer vision and machine learning could address what he calls “the paperwork pyramid” at the base of healthcare. Eligibility checks, faxes, and prior authorizations consume much of the industry’s estimated $450 billion in annual administrative costs. Lightspeed led a $2 million seed round to build Coral, and Singh joined the board as an advisor. “Ajay had both deep technical skills and unusually high emotional intelligence,” Singh said. “He understood that fixing healthcare meant working within its complexity, not against it.”

Shrihari says Coral’s team learned by embedding with clinic staff and mapping every step of the intake process. “You can’t make a mistake in healthcare,” he told me. “Anything below the high 90s just isn’t usable.” He adds, “We don’t replace people. We design the system to work with them because their judgment is part of why care gets delivered safely.”

Traditional robotic-process-automation (RPA) tools struggled in clinics because generic optical-character-recognition (OCR) systems and fragile browser bots failed to read handwriting, flipped insurance cards, and blurry text. Coral trains its models specifically on healthcare data, using real-world clinic documents instead of synthetic samples. Its platform adapts to the irregularities of each provider’s workflow, keeping humans in the loop for oversight.

This is a crowded market, with legacy RPA vendors like UiPath and Automation Anywhere serving large systems, and new entrants chasing niches such as infusion, radiology, and pharmacy. Coral differentiates itself by accepting the messy status quo and automating inside it. “Legacy vendors tried to rebuild everything,” Shrihari said. “We integrate with what providers already use so value appears on day one.”

DASCO, a home medical equipment provider offering respiratory therapy, mobility aids, to patients, is one of Coral’s early customers. “Their AI-driven software streamlines intake of documents with enhanced accuracy and speed, reducing turnaround times from hours/days to mere minutes,” Said DASCO President Michael Gorman, in an email. “This not only alleviates a significant administrative burden on providers but also benefits countless patients.”

A July 2025 MIT report, The GenAI Divide, found that 95 percent of organizations saw no measurable business return from generative AI pilots, with only five percent showing meaningful results. While the study briefly shook tech stocks, critics argued it ignored smaller firms and individual users—those closest to real operational pain points. Coral AI’s story illustrates that point.

“If The GenAI Divide captures widespread disappointment in corporate AI spending, Coral shows where success is emerging,” Singh said. “Once you can automate intake and referral at this level of reliability, you can apply the same methodology to any industry with unstructured workflows: insurance, banking, even government. The foundation they’re building has range.”