Mercor

Paying a physician, a lawyer or an accountant to teach a machine how their job is done is the idea Mercor built a company around. The company runs a network of experts who are hired to train and grade artificial intelligence models, and it reports that more than five million people are in that network. Mercor's pages put the average contracted rate at $90 an hour and say daily payouts to experts pass $4 million.

Building the layer between human expertise and frontier models

Mercor was founded in 2023 by Brendan Foody and Adarsh Hiremath, who describe the bottleneck in AI as human intelligence rather than computing power. Its work for AI labs centers on human data and evaluations: benchmarks and reinforcement learning environments built from real professional tasks, graded by people who do those jobs. The company says its network covers more than 300 professional fields and that candidates are vetted with role-specific AI interviews. In July 2026 it announced the acquisition of Deeptune, a team that recreated hundreds of enterprise applications so that agents can practice professional workflows.

APEX benchmarks and partner leaderboards

Mercor publishes a family of benchmarks under the APEX name, with public leaderboards for agents, accounting and software engineering. It also builds partner benchmarks with outside organizations, among them a legal benchmark of 1,749 attorney-built tasks across 24 practice areas, an engineering benchmark made with Google Cloud AI Research, and a banking customer-support benchmark from Sierra. The enterprise pages add data monetization, in which a business licenses anonymized workflow data, with automated masking of more than 60 categories of sensitive identifiers.

Enterprise agents with forward deployed engineers

For companies deploying AI in production, Mercor offers diagnostics that map workflows and rank agent opportunities, deployment of custom agents by forward deployed engineers, expert-graded evaluations, and a platform that unifies context, connectors and evaluation. Its engagement summaries describe an incident response agent that cut median time to mitigation by more than half and a support system handling about 12,000 requests a week. These figures are the company's own.

What the Series D post says

In a post dated 10 November 2026, chief executive Brendan Foody announces a $750 million Series D that values Mercor at $20 billion. The post says the round included an employee tender and names investors including Benchmark, Felicis, General Catalyst, NVIDIA, Salesforce and Sapphire Ventures. It adds that the company opened offices in New York and London and will use the money for evaluation and post-training infrastructure and for more environments where agents learn complex workflows.

Key facts

Founded: 2023 Headquarters: San Francisco, California (registered provider in its terms: Mercor.io Corporation, 181 Fremont St., San Francisco, CA 94105) Offices: San Francisco, New York City, London People: Brendan Foody and Adarsh Hiremath, co-founders and co-chief executives Products or Services: Human data and evaluations, APEX benchmarks, expert hiring network, enterprise AI diagnostics and agent deployment, data monetization Funding stated by the company: $750 million Series D at a $20 billion valuation, announced 10 November 2026, including an employee tender; earlier $350 million Series C at $10 billion Website: mercor.com