The Company

Skan was founded in 2018 by CEO Avinash Misra and COO Manish Garg, who had sold their previous start-up, mobile solutions vendor Endeavour, to Genpact in 2015. After completing an undisclosed seed round in January 2020, Skan went on to raise a $14 million Series A in October 2020 led by Cathay Innovation, with participation from Citi Ventures and Bloomberg Beta. A $40 million Series B followed in March 2022, led by Dell Technologies Capital and joined by fellow new investors Liberty Global Ventures, Firebolt Ventures, Zetta Venture Partners, and GSR Ventures (along with repeat participation from Cathay Innovation and Citi Ventures).

With over 100 employees, Skan is headquartered in Menlo Park, California, with development situated in Bangalore, India, and additional offices in Seattle, Washington, and Ottawa, Canada.

The Technology

Skan offers what it calls a “Business Process Intelligence Platform”; a combination of process and task mining but without the integration work often required for the former. This allows them to generate a data model with the end-to-end view of processes but with full detail of all the manual effort and costs, underpinned with an analysis, reporting layer, and operational management platform (see Figure 1). This supports where Skan believes the market is currently heading: from the discrete discovery and analysis of processes to an integrated set of methods enabling organizations to continuously transform their ways of working with real-time operational control and monitoring for improvement.

At the data collection, Skan employs “Virtual Assistants (VAs)” to capture first-party clickstreams, application data, and screen capture information at an employee desktop level. This approach allows Skan to avoid difficult or time-consuming integration tasks to access application logs, instead collecting natively at the point where the workforce accesses those applications. This means that Skan starts any engagement with a blank slate; however, the data set builds up rapidly when the VA desktop agents are installed across a significantly sized workforce, thus providing a consistent and reliable stream of information across multiple simultaneous applications on which process discovery can be based.

Like other similar approaches, this form of data capture is currently limited to Windows desktops, although Skan is among those looking to expand to other desktop and mobile operating systems, in time.
To keep this potentially rich data stream secure and anonymous, and only let it reveal what is considered acceptable for analysis purposes, the Skan architecture is set up to make sure that clients can retain sensitive data storage within their own network, which includes client-specified data types collected by the desktop agents. This data can be collated and analyzed within a client’s own DMZ before routing only the anonymized metadata to Skan’s AI engine and analytical data store, outside the client network. Other approved processing application data can route straight to Skan as required, and both can be reconciled for analysis.

Within Skan’s analytical data store, that information is contextualized with a transaction identifier (if applicable) – process, sub-process, and task, (e.g., looking up customer record, checking customer contacts) – so that the elements within the individual desktop clickstreams can be presented in a unified view across functions and organizational levels. Each can be individually analyzed, updated where required, and applied to the categorization models being employed client-by-client, to improve ongoing accuracy.

This ability to apportion individual clicks to likely intents and purposes further allows Skan to deduce which sets can be associated with a process/task, and which of these tasks can be attributed to an element or iteration of a process. From this data, processes themselves and their variants (alternative sets of interactions designed to conclude with an identical or adjacent end point) can be established. By capturing identifiable elements within the data (“attributes” such as case IDs), complete value chain lifecycles can additionally be deduced.

This set of activities outputs graphically modeled depictions of the normalized processes – in essence, process maps (see Figure 2). Combined with their history of execution and versions of the derived or documented reference process, Skan refers to these maps as the “Digital Twin of Operations.” This is an awkward term for useful output: a model of a process that can be contextualized within the platform as to how it fits into the broader operations of the organization, and actionable recommendations. In addition, patterns for potential suitability for automation can be identified (e.g., likely trigger points, mix of structured and unstructured data). Furthermore, scoping of automations can be done in two dimensions (which steps, for which transactions) and this can reduce risk while decreasing elapsed time, as a long tail of exceptions can be excluded. The tool also has standardized views for quantifying benefits realized.

That “Intelligent Automation Discovery” framework is just one framework that Skan can apply to process data it has captured, and Skan believes this is what primarily distinguishes its offering. Other elements are statistical control (exception management, based on expected metrics), process health (benchmarking against common KPIs, such as wait time) and operations control (continuous visibility over workforce operations). These frameworks (“value engineering,” as Skan describes them) are designed to ensure that the platform can be deployed in a continuous, operationalized manner from the get-go, with those anticipated value-added use cases available immediately.

That such operational deployments are intended to be available does not mean clients have to operationalize if they aren’t ready or don’t want to. Skan’s current projects are generally operated in three phases. Initial discovery projects – normally discrete – are generally Skan-led and small-scale value-proving operations. They lead to a more equal customer/vendor partnership in secondary phases where the platform can be directed toward repeatable work across a wider range of operations. Ongoing customer-led – but still Skan-supported – operational phases are the ultimate goal, and these can take several years to fully realize for enterprise customers (generally driven out of customer success or solutions delivery lines of reporting).

Skan currently has nascent clusters of customers in four areas where it is focusing strategically: insurance, healthcare, banking, and financial services. Use cases from customer onboarding to claims management are common initial projects for Skan adoption in these industries, with an emphasis on efficiency and effectiveness driving the primary approach. Other verticals where Skan has a toehold include manufacturing and high technology. Skan has technical partnerships with automation vendors UiPath and EvoluteIQ, which makes sense given that Skan’s platform offers valuable suitability insights for the use of such tools and the company has roadmap plans to simplify the process of working with automation platforms.