Harbor Wants to Kill the Eight-Week EDC Setup That Slows Down Every Clinical Trial

The Macro: Clinical Trial Data Management Is Painfully Outdated

Running a clinical trial is one of the most data-intensive operations in any industry. Every patient visit generates forms, measurements, lab results, and observations that must be captured, verified, and stored in a system that meets FDA requirements. The software that manages this data is called an electronic data capture system, or EDC.

The problem is that EDC systems are ancient. The market leaders, Medidata Rave and Oracle’s InForm, have been around for decades. Setting up a new study in these systems takes months. You need database programmers to build custom forms. Every field, validation rule, and edit check must be configured manually. The setup cost for a moderately complex trial can run into hundreds of thousands of dollars.

And that is just the setup. Once the study is running, research site staff spend hours every day manually entering data from source documents into the EDC. This is tedious, error-prone work that pulls clinical staff away from patient care. Monitors then visit sites to verify that the data in the EDC matches the source documents, a process that costs sponsors millions in travel and staff time.

The entire workflow is built on assumptions from the pre-AI era: that humans must build forms by hand, that humans must type data from paper into computers, and that humans must physically check that the typing was done correctly. Every one of those assumptions is now questionable.

Harbor, backed by Y Combinator, is building an AI-native EDC that automates each of these steps.

The Micro: Protocol In, EDC Out, Ten Minutes

Harbor’s flagship feature is Magic Build. Upload a clinical trial protocol PDF, and Harbor automatically generates the electronic case report forms and source documents in under ten minutes. The traditional process takes eight weeks. That compression alone would justify switching for many sponsors.

Magic Capture uses AI to extract data directly from source documents into the EDC system, reducing manual data entry by what Harbor claims is 90%. The AI reads the source, identifies relevant data points, and populates the EDC fields. Research site staff review and confirm rather than type.

Magic Monitor provides risk-based remote monitoring with confidence scores for each extraction. This enables centralized monitoring without requiring monitors to travel to every site and compare paper documents to screen entries. The time and cost savings for sponsors are substantial.

The platform supports ePROs for patient-reported outcomes, unlimited source document storage, and full 21 CFR Part 11 compliance with immutable audit trails and electronic signatures. These are table-stakes requirements for clinical trial software, and Harbor covers them.

The founding team has direct domain expertise. Albert Cai previously led clinical trials and regulatory strategy at Biolinq, helping advance a product through FDA De Novo authorization. Nathan Leung has five years of tech experience including time at Ramp and was the first employee at a previous YC startup.

Pricing is structured for accessibility: free for academic and non-profit studies, starting at $2,000 per month for commercial studies, and custom pricing for enterprise CROs. The free academic tier is a smart adoption strategy because academic researchers are often the earliest adopters of new clinical trial technology.

The competitive space includes Medidata (owned by Dassault Systemes), Oracle Health Sciences, Veeva Vault EDC, and Castor. Among these, Castor is the most modern and user-friendly, but none of them offer the kind of AI-native features Harbor is building.

The Verdict

Clinical trial data management is a multi-billion-dollar market trapped in the past. Harbor’s AI-first approach could genuinely reshape how trials are run if the accuracy holds up under regulatory scrutiny.

At 30 days: how many studies are running on Harbor, and have any passed FDA audit? Regulatory acceptance is the ultimate validation for clinical trial software.

At 60 days: what is the data accuracy rate for AI-extracted data compared to manual entry? If accuracy meets or exceeds human data entry, the value proposition is clear.

At 90 days: are CROs adopting Harbor across multiple studies, or is adoption limited to individual sponsors? CRO adoption is the scale play because CROs manage trials for dozens of sponsors simultaneously.

I think Harbor is going after one of the biggest opportunities in healthcare IT. The legacy EDC vendors are ripe for replacement, and the AI-native approach addresses the three biggest pain points: slow setup, tedious data entry, and expensive monitoring. The free academic tier should build a strong base of early users who then advocate for Harbor when they move to commercial sponsors.