Skan AI raises $63 million betting that watching how employees actually work is the missing layer of…

What happened

Skan AI has closed a $63 million Series C round co-led by Cathay Innovation and Dell Technologies Capital, a signal that enterprise investors are willing to back an approach to workplace intelligence that sits below the layer of what employees say they do and above the layer of raw system logs. The company's core premise is that most enterprise AI initiatives fail not because the models are wrong but because the training and context data feeding them is wrong — built from process documentation, interviews, and ticket systems rather than from observed work as it actually happens.

The product Skan AI builds is described as a context graph of work. That term implies a structured, queryable representation of how tasks move through an organization: which applications are touched in what sequence, how long each step takes, where humans hand off to automated systems and where they take back control, and what patterns of behavior repeat across teams or diverge between individuals doing nominally the same job. Building this requires continuous observation of employee desktop activity across the suite of enterprise software a company already runs — not instrumented APIs, not surveys, but screen-level behavioral capture that is then abstracted into process-level signals.

The technical detail

For engineers and builders, the architectural bet here is worth examining. Most process mining tools instrument at the system layer, reading event logs from ERP or CRM platforms. That approach gives you clean timestamps but only captures events the software itself surfaces. Skan AI's approach, capturing at the UI or screen level, means it can observe work that crosses multiple systems in a single human action, catch workarounds that employees build outside of official workflows, and detect latency that exists in human cognition and context-switching rather than in system response time. The resulting graph is potentially richer than log-based alternatives but also more computationally and privacy-sensitive to collect and process.

Why it matters for builders

The enterprise AI market has spent several years grappling with the problem that AI agents perform poorly when the workflows they are meant to automate were never precisely defined in the first place. Skan AI is making an explicit bet that a layer of work observation has to precede deployment of agentic systems — that you cannot send an AI agent into a process you do not understand at behavioral resolution. The $63 million suggests that Cathay Innovation and Dell Technologies Capital share that view, or at least believe the framing is compelling enough to win enterprise budget from buyers who have already committed to AI transformation roadmaps and are looking for foundational infrastructure.

The competitive context for this raise is notable. Process mining as a category has a dominant player in Celonis, which has raised well over a billion dollars and built its business on ERP log analysis. Automation players like UiPath have also moved into process discovery as a natural complement to their RPA tooling. Skan AI's differentiation appears to be the level of behavioral granularity it captures and its framing around AI readiness rather than pure process optimization. Dell Technologies Capital's participation is also worth noting — Dell has a direct interest in enterprise software infrastructure and in tools that make AI deployments inside large organizations more legible and controllable.

Market and competitive context

The practical question coming out of this raise is how Skan AI converts observational data into something that enterprise AI buyers can act on at speed. Building a context graph is a data engineering problem; translating that graph into improved AI agent outcomes or measurable process efficiency is the harder product problem. Buyers will want to see time-to-value measured in weeks, not quarters, and they will want the system to surface specific intervention points rather than produce dashboards that require analysts to interpret. The company's Series C scale suggests it has enough runway to push into that product maturity, and the involvement of Dell creates a natural distribution channel into large enterprise accounts.

What to watch next

The open questions are real. Employee observation at the level of screen capture raises legitimate concerns that any serious enterprise deployment will have to address — data residency, what is retained, who can query the graph, and whether employees have meaningful transparency into what is being collected. Regulatory environments in Europe and in heavily unionized industries will create friction. There is also a model fragility risk: if the context graph is built on observed behavior during a period of transition, it may encode inefficiencies rather than best practices, and AI systems trained on that graph inherit those patterns. Whether Skan AI has built mechanisms to distinguish between descriptive capture of how work happens and prescriptive guidance toward how it should happen will be a key technical and commercial differentiator as it scales into the enterprise accounts this round is presumably meant to unlock.