Problem

Mercor addresses the bottleneck that advanced AI systems cannot improve reliably without high-quality human judgment, domain expertise, and feedback. Models still struggle with messy real-world workflows, the tools and applications people use, and recognizing what good work looks like in a particular profession or business. For AI labs, this creates a costly, labor-intensive need for expert-generated training, evaluation, and reinforcement data; for enterprises, it makes it difficult to move AI agents from generic demonstrations into economically useful production work. The scale of the pain is substantial: Mercor has reported paying more than $1.5 million per day to people training AI systems.

The central use case is turning scarce professional expertise into usable AI-development data. Lawyers, doctors, engineers, finance professionals, creators, and other specialists evaluate model outputs, complete realistic tasks, and provide the judgments and examples needed to improve models and agents. For enterprises, the same capability helps capture how employees actually work so that agents can be trained against real workflows rather than abstract simulations.

Product / Service

Mercor operates a two-sided marketplace and AI-training infrastructure. It connects professionals with remote projects for leading AI companies, uses an adaptive AI interview to assess expertise at scale, matches approved experts to scoped assignments, and handles payments to contractors across countries and currencies while billing AI labs and enterprise customers. Experts work on tasks such as evaluating, improving, and guiding large language models, with compensation and project terms defined up front.

The company is expanding beyond labor placement into enterprise agent deployment. Mercor says it learns how work gets done inside a company, builds agents equipped with human expertise, and defines what good performance looks like. Its resulting datasets, benchmarks, evaluation criteria, and workflow traces can help labs improve models and help businesses deploy agents with quality controls, while real-world usage feeds new edge cases and performance gaps back into the training process.

Market

Mercor competes across AI data and model-training infrastructure, expert networks, AI-enabled recruiting, model evaluation, and enterprise agent deployment. Its closest named rivals in the AI data race include Scale AI and Surge AI, although Mercor’s positioning increasingly spans both the supply of expert human work and the deployment of agents trained on that expertise. The company’s differentiation is the combination of a large vetted expert network, automated assessment and matching, payments infrastructure, and proprietary data about economically valuable workflows.

Mercor is not pre-revenue. The company reported a $350 million Series C that brought its valuation to $10 billion, and its own materials describe it as a profitable Series C business trusted by six of the seven major technology companies commonly called the Mag 7. Secondary reporting says it reached $2 billion in annualized gross revenue in June 2026, after reaching a $500 million run rate in 17 months. Mercor also says its network includes more than four million vetted experts and 1.9 million referrals, indicating substantial marketplace traction and a potentially reinforcing data and supply-side network effect.