Avinash Misra and Manish Garg have raised $63 million for Skan AI, backing a bet they began in 2018: enterprise agents need to learn how employees actually work before they can reliably replace or assist them.
The Series C, reported by VentureBeat on August 12, was co-led by Cathay Innovation and Dell Technologies Capital. Citi Ventures, Bloomberg Beta, State Farm Ventures and Wipro Ventures also participated. VentureBeat reported that the financing brings Skan AI's total capital raised to roughly $120 million. The company's valuation and other deal terms were not disclosed.
Previously disclosed backers include Zetta Venture Partners, Firebolt Ventures, Plug and Play, GSR Ventures and Liberty Global Ventures, according to Skan AI's Series A and Series B materials.
Misra and Garg have been building together since their student years. They grew up together in Kanpur, India, attended the Indian Institute of Technology Kanpur and previously co-founded Endeavour Software Technologies, an enterprise mobility provider. Genpact agreed to acquire Endeavour in September 2015; the acquisition price was not disclosed.
Their second company grew from what they saw during enterprise transformation projects: process documentation described the approved procedure, while employees handled the real work through spreadsheets, emails, legacy applications, judgment calls and improvised fixes. The company says it was founded in 2018 after Misra and Garg saw a gap between documented procedures and the way employees actually carried out enterprise work.
"Why is it so difficult for an organization, and a large enterprise especially, to understand how its own work actually gets done?" Misra told VentureBeat. "Why does it need to fly in McKinsey consultants for that?"
Turning observation into agent instructions
Skan AI installs an observation layer on employee desktops and follows work as it moves among approved applications. In its description of Skan AI Intelligence, the company says the software captures clicks, application switches and handoffs, maps workarounds and exception paths, and can observe activity across legacy systems, mainframes, Citrix and virtual desktop environments, desktop software and web applications. In a separate company explanation of its screen-level technology, Skan says its workstation agents do not depend on APIs or system logs.
Skan AI then abstracts those observations into what it calls a "Context Graph of Work." The central technical claim concerns intent extraction: recognizing whether a worker has started a new case, corrected an old one, escalated an exception or followed an undocumented shortcut. Correctly translating a stream of interactions into a stateful business process is the harder problem.
"The hard problem is not screen observation," Misra told VentureBeat. "The hard problem is abstraction of what you see on the screen - the intent extraction."
The new financing arrives with the general availability of two products. Skan AI Blueprint gives executives a map of potential AI projects, including readiness scores, estimated financial impact and ROI tracking. In its description of Skan AI Agents, the company says the product uses observed process models, which it calls Agent Operating Procedures, to execute enterprise workflows. Skan AI Intelligence, the existing analysis layer, maps process variants, bottlenecks and automation candidates.
Together, the products push Skan AI from process discovery into the larger budgets forming around agent deployment. Blueprint helps decide where automation spending should go. Intelligence models the work, and Agents executes it. Skan can sell the planning, data and execution layers as one system rather than remain an analytics product feeding another automation vendor.
The timing reflects weak returns from many corporate AI projects. VentureBeat reported that Gartner research cited by Skan found only 8% of enterprises had AI agents in production and projected that 95% of early implementations would require a complete redesign. Skan's pitch is that agents fail when they receive official procedures and backend records without the undocumented decisions that shape daily execution.
That expansion raises the cost of an inaccurate model. A flawed process map produces a weak dashboard. An agent using that map can make the wrong decision, mishandle an exception or act inside a regulated workflow without the required evidence.
Company-reported traction
Skan AI claims it has observed more than 500 processes and delivered more than $500 million in value. Those figures are company-reported and unaudited, and Skan has not publicly explained enough of its methodology to independently assess the value calculation.
A March 2022 company funding announcement said more than 20 major financial-services, healthcare and insurance companies used its technology. That was a historical disclosure rather than a current customer count.
In a February 2024 case study about a Fortune 500 financial-advisory firm, Skan said the deployment expanded from a 25-to-30-user proof of concept to approximately 200 users and identified at least 20 employee-experience and cost-reduction opportunities. Skan did not name the customer, and the results remain company-reported.
A crowded race to provide enterprise context
In VentureBeat's account, Misra drew a distinction between backend records and the work that precedes them. System logs show completed transactions, he argued, while employees review documents, move among applications, consult colleagues and apply institutional knowledge before committing those transactions. Skan's screen-level observation is meant to reconstruct that interval.
"All backend data, by definition, is a committed state of work," Misra told VentureBeat. "Work is really what happens between those committed states."
Skan AI does not have that market to itself. On its Context Model page, Celonis describes the product as a dynamic, real-time digital twin of operations that supplies operational context to people and enterprise AI. UiPath says its Process Intelligence offering supplies process data for workflows involving orchestration, software robots, AI agents and employees. Mimica says it records desktop actions and identifies process-improvement and automation opportunities. Soroco says its Scout software captures digital interactions and reveals workflow variations and automation opportunities. Gartner Peer Insights lists Celonis, UiPath, SAP Signavio and IBM Process Mining among the alternatives considered by Skan users.
Skan AI's position depends on the quality of its abstraction, the process data accumulated since its 2018 founding and its ability to deploy inside security-sensitive enterprises. Skan argues that its stated coverage of mainframes, virtual desktops and legacy software can address processes for which backend records are incomplete, although desktop-observation rivals can pursue the same gap. The $63 million gives Misra and Garg additional time to convert that technical claim into production deployments before larger automation platforms close the distance.
Cathay Innovation and Dell Technologies Capital have backed Skan AI before. Cathay led a $14 million Series A announced in October 2020. In a March 2022 funding announcement, Skan said Dell Technologies Capital led its $40 million Series B. Their decision to co-lead the Series C is a renewed bet on the founders as Skan moves deeper into execution software.
Observation brings a surveillance test
Skan AI's premise creates an unavoidable workplace question: software that watches screens to understand processes can also be used to monitor people.
Skan says its controls aggregate behavior across groups, limit observation to approved applications and URLs, and keep the resulting data behind the enterprise firewall. Those are company-reported architectural claims, not findings from an independent audit.
Misra's wager is that enterprises will accept that scrutiny because the operational data is valuable. General-purpose agents arrive with little knowledge of how a particular insurer processes a claim or how a bank handles an unusual compliance case. Skan AI wants to own the layer that teaches them.
The Series C funds a shift from watching work to acting on it. Misra and Garg spent years building the observation system before enterprise agents became a board-level priority. They now have to prove that process knowledge gathered from employees can produce agents that are accurate, auditable and trusted by the people whose screens supplied the underlying data.