Skan AI raises $63M to build a context graph of work for enterprise AI agents

Skan AI raised $63 million in a Series C round to build what it calls a "context graph of work" for enterprise AI, a layer that records how work actually gets done across applications and feeds that record into AI agents. The round, co-led by Cathay Innovation and Dell Technologies Capital with participation from Citi Ventures, Bloomberg Beta, State Farm Ventures, and Wipro Ventures, brings total funding to roughly $120 million since 2019.

What Skan AI actually does

The company deploys software on employee desktops that captures screenshots and processes them locally. The images never leave the machine. What gets sent to Skan AI's analytics platform is anonymized metadata: which applications were used, in what order, and where decisions were made. This lets the company map operations without collecting work product.

Skan AI splits the platform into three products. Blueprint maps how a process moves across systems and teams. Intelligence pinpoints where time and money leak out of that process. Agents deploys automation modeled on what a company's top performers actually do.

Why this matters for enterprise AI builders

Most enterprise AI projects fail because agents lack real operational context. They operate on documentation, system logs, or stale process maps that don't reflect how work actually flows. Skan AI's approach builds that context from direct observation, which Cathay Innovation partner Simon Wu described as becoming an infrastructure layer for corporate AI in the way CRM became the system of record for customer data.

For builders shipping AI agents into regulated enterprises, the on-device processing model is worth noting. Keeping screenshots on the user's machine and sending only anonymized metadata addresses a common data governance objection. It doesn't eliminate all privacy concerns, but it makes the pitch easier for compliance teams.

What the numbers look like

Skan AI reports revenue growth of more than 300% year over year, net dollar retention of 150%, and over 25 billion work signals logged. Customers include a quarter of the Fortune 50, seven of the 10 largest U.S. banks, and three of the five largest U.S. insurers. U.K. facilities management group Mitie is also a customer.

In one deployment at a large U.S. bank, Skan AI tracked 11.2 million context switches across 1,500 finance staff and surfaced $37 million in operational friction. Agents built on those observations cut cost per transaction by 32% and lifted throughput by 41%, worth $18 million in annualized savings, according to the company. Across its customer base, Skan AI puts average operational savings at 30% to 40% and cumulative measured customer value at more than $500 million.

What to watch for

All performance metrics and customer results cited here are vendor claims as presented in Skan AI's press materials and the SiliconANGLE report. Independent verification of the bank deployment numbers or the claimed savings is not available in the provided sources. The company plans to spend the new funds on product development and deeper go-to-market efforts in financial services, insurance, healthcare, and technology.

For builders evaluating similar approaches, the key question is whether the context graph generalizes across different enterprise software stacks and whether the agent models trained on observed behavior can handle edge cases that top performers don't encounter regularly. Skan AI's customer concentration in banking and insurance also means the approach may be better validated in highly structured, compliance-heavy workflows than in more fluid knowledge work.