Crux Labs raises $6M to scale human judgment for AI agents

Crux Labs emerges from a specific operational gap: enterprise AI agents can generate thousands to millions of daily interactions, but existing observability platforms mainly help humans monitor and inspect traces rather than improve agent behavior at scale. As a result, many teams sample the top 1% or less of traces based on a metric, threshold, or random selection, so the vast majority of agent failures go unnoticed. The company argues that if 'what good looks like' stays in experts' heads, fixes arrive late—sometimes weeks after a small glitch escalates to a ticket and a patch. Crux's approach uses its own agents as evaluators to capture employees' expert judgment and calibrate it across millions of interactions, aiming to reduce that cycle by putting experts and AI agents on the same level. Co-founder and CEO Julie Kheyfets framed the challenge as aligning AI agents to human judgment, while CTO Krishna Kulkarni described enterprises as lacking a bridge between engineering teams who build agents and subject matter experts who know success. The company says it has seen strong traction across high-growth startups and Fortune 500 businesses in healthcare, cybersecurity, and financial services.