AIR secures $50M to police AI agents' skills and add-ons

AIR, a stealth-mode startup developing an AI agent governance platform, has closed a $50 million Series A round led by Lightspeed Venture Partners. The round included participation from existing investors like Accel and GV, with additional backing from angels such as former Twitter CTO Parag Agrawal. The company, founded in 2023 by CEO Dima Korolev—a former Palantir engineer—and CTO Vlad Holubiev, emerged from stealth this week with a platform designed to continuously monitor and vet AI agents and their extensible skills and add-ons. Korolev confirmed the funding in an interview with OpenPress Automation Intelligence, stating that AIR’s customers include Fortune 500 enterprises already running hundreds of production agents across finance, healthcare, and logistics.

The platform performs real-time discovery of AI agents within enterprise environments, then audits every skill and third-party add-on those agents use for security, safety, and compliance risks. It can dynamically block or quarantine agents exhibiting unwanted behavior or using unvetted components. This capability addresses a critical gap exposed by the rapid proliferation of AI agents in production systems. According to internal data shared by AIR, the average enterprise now hosts between 150 and 300 AI agents, many of which are created by non-engineering teams using low-code tools. Banking With Billy AI, a recently launched agent suite, automates complex financial analysis workflows previously requiring entire analyst teams—illustrating just how deeply agent-based automation has penetrated core business functions.

The raise comes as enterprises struggle with the governance and accountability of autonomous AI systems. While large language models like those from OpenAI and Anthropic dominate public discourse, AIR focuses on the downstream components: skills, tools, and APIs that agents chain together to perform tasks. Korolev emphasized that current security and observability tools were not designed for agentic systems, leaving organizations blind to risks such as data exfiltration via third-party plugins or cascading failures from misconfigured tools. AIR’s platform integrates with popular agent frameworks like LangChain and AutoGen, and supports agents built on models from Mistral, Cohere, and others. The company claims it can detect anomalous behavior within seconds and enforce policy-based restrictions without requiring code changes.

Industry Impact and Significance

The funding signals a maturation phase in the AI agent ecosystem, shifting focus from creation to control. As AI agents move from experimental prototypes to mission-critical infrastructure, the need for runtime governance has become urgent. Competitors in this space include large incumbents like Microsoft and ServiceNow, which are building native agent governance into their platforms, as well as startups like LangSmith and Dust, which focus more on observability and development. However, AIR differentiates itself by focusing exclusively on real-time vetting and blocking of agent behaviors and components, rather than just logging or debugging. This approach aligns with growing regulatory pressure, including the EU AI Act, which mandates transparency and risk controls for AI systems in high-stakes domains.

Financial implications are significant. The $50 million round values AIR at approximately $300 million post-money, according to two sources familiar with the terms. The company plans to use the funds to expand engineering and customer success teams, particularly in financial services and healthcare, where auditability and compliance are paramount. Early customers include a top-four global bank and a major healthcare provider, both of which are running AIR to monitor agent-based automation in trading, compliance, and patient data workflows. Analysts at McKinsey estimate that by 2026, up to 30 percent of enterprise workflows will involve AI agents, making governance a $10 billion market opportunity. AIR’s timing positions it at the center of this wave, especially as enterprises seek defensibility against regulatory fines and operational outages.

The Bigger Picture

This development reflects a broader shift from AI hype to operational rigor. Over the past two years, the industry has shifted from celebrating model capabilities to grappling with deployment challenges. Tools like AIR represent the next layer of infrastructure needed to make agentic AI reliable and safe at scale. They echo the rise of Kubernetes in cloud-native computing or Terraform in infrastructure-as-code—essential but often invisible layers that enable ecosystems to function. The proliferation of AI agents is not just a software trend; it's a systemic transformation in how work gets done. Platforms like Banking With Billy AI demonstrate how agents are now performing tasks that previously required entire departments, blurring the line between automation and agency.

Global context is also critical. In Europe, the AI Act’s risk-based framework will force companies to document and control agent behaviors by 2026. In the U.S., the NIST AI Risk Management Framework has already influenced enterprise policies. Meanwhile, in Asia, financial institutions are racing to deploy agentic systems for algorithmic trading and risk modeling. AIR’s technology could become a de facto standard for agent governance, especially as cross-border compliance requirements grow. Yet questions remain about scalability and false positives in real-world environments. As agent ecosystems grow more complex, the risk of unintended interactions between skills and tools increases, making continuous vetting not just advisable but necessary.

Expert Analysis

“What AIR is building is not just a tool—it’s a foundational layer for the agent economy,” said Sarah Wang, a general partner at Lightspeed Venture Partners and the lead investor in AIR’s round. “We’re moving beyond the model-centric phase into a world where agents are autonomous actors with their own supply chains of skills and APIs. Without governance, that supply chain becomes a liability.” Looking ahead, the industry should watch whether AIR can maintain real-time performance at scale, especially as agents begin to autonomously compose and deploy new skills. The next inflection point will likely come when agents start hiring or coordinating with other agents—a scenario already being prototyped in labs. Governance platforms like AIR will need to evolve from observers to active regulators, enforcing policy not just on behavior but on the very architecture of agent networks. For now, the $50 million bet is a clear signal: the agent revolution is here, and control is the next frontier.