The Company

Founded in 2018 and headquartered in Cologne, Germany (with additional development resources based in Tychy, Poland), KYP.ai was created by a group of ex-Capgemini employees. Initially, it focused on discovering and recommending business-case-driven improvement opportunities for process automation and transformation. The company raised a seed round for an undisclosed amount in September 2021 from Tola Capital, 10x Group, and 42CAP.

As the use cases of the technology grew to include improving workforce management, employee engagement, and the customer experience, the vendor increased its focus on the wider range of insights that it could offer. A pivot toward holistic process analysis for productivity mining followed in 2022. The company currently employs over 50 people in its two European locations and a sales office in the US.

The Technology

KYP.ai saw the potential of the insights it generated for its clients and that they went far beyond finding opportunities for automation. The analysis of single-user paths quickly told KYP.ai that finding a commonality among groups of users attempting a common path provided a richer process map on which to build. Accordingly, KYP.ai decided to focus on providing a detailed analysis of processes where its insights are reliable because of how process data was collected. (See Figure 1 for an example of an analysis of invoice processing.)

Put another way, KYP.ai goes beyond process mining and intelligent automation and generates insights for a range of additional opportunities, including improving process efficiency, workforce management in the hybrid home/office model, employee engagement and work/life balance, end-user and customer experience, system performance, and, most recently, process and enterprise sustainability.

KYP.ai collects raw data by installing an agent application on user computers that assembles two primary forms of information. First, screen reading (e.g., URLs, query strings, text labels) sends back textual information about each task. Second, full-screen captures are taken and sent to the platform for analysis. The two collection methods can be deployed to collect data on use of desktop applications that have been agreed with the client, for example ERP, and excluding others such as employees’ personal email or banking apps. This approach reduces noise in the returned data and provides an initial privacy point; the second is that the platform can anonymize team member identities, leaving no personal data present in any subsequent analysis. The agent application is currently only available for Windows, but KYP.ai told us a macOS option is under development.

In operation, KYP.ai generates a significant volume of data to convert into something that represents common user paths through processes. KYP.ai uses AI (natural language processing along with a machine-learning-derived classifier) in combination with a rules engine to normalize the data. Using data points from each step in the process, KYP.ai then classifies what is occurring via a classification engine it developed using common application signifiers. For example, when a new user employs a common SaaS application, the signifiers found in that step will likely have been previously registered and classified (for example) as a type of order process or customer record. KYP.ai tells us that its classification engine has been trained on 100 billion records and provides the engine as a general model to all its customers.

What a process produces in KYP.ai are essentially clusters of steps broadly classified as being used for the same general purpose, rather than subtly different but functionally identical approaches. This helps speed up the time it takes for a process to be identified and analyzed. In addition, transaction IDs can be captured in order to analyze specific tasks or customer cases.

A rules engine is added to the AI classification method; it can be applied equally to text- and image-collected data, thus allowing manually created rules based on customer-specific patterns or paths through the process or identification of custom applications that might fall outside the learning of the AI. Rules can be stacked in order and regular expressions applied for more complex identification and processing.

Image analysis can be used for data collection where the applications provide more visual detail through a clickstream. For instance, in the straight text collection method, thick client applications such as Microsoft Excel or those that use right-click and context-sensitive menus to guide users may be hidden from view. Hence, KYP.ai recommends that images are collected in these situations and then processed by the agent application. The company suggests that an impressive sub-20-millisecond processing time is achieved for this type of analysis.

The output of the collection methods and post-normalization processing are detailed visual representations of the processes taking place within the organization. These are broken down into a customer-defined organizational structure based on where individual agent applications fit into a team structure. This normalization ensures that paths can, to a large degree, be automatically annotated to be easily explained. Manual augmentation to add customer-specific information is also possible within the platform.

The company is currently helping to understand customer service pathways, supply-chain throughput analysis, and comparative productivity analysis for remote, home-working, and office-based employees. KYP.ai sells directly to organizations (both SaaS and on-premises) and has white-labeling and partnership agreements with other technology providers including AYR.ai, HuLoop, XpertRule, and the consultants Emergence Partners. It is a member of the Intelligent Automation Collective (IAC) that includes AntWorks, Enate, OpenDialog, and RPA Supervisor as well as XpertRule and Emergence Partners.