The Company

KnowledgeLake was founded in 1999 by
Ron Cameron and Bob Bueltmann. Cameron remains the CEO today. The company started out as a system integrator and reseller of products like FileNet, Datacap, and Kodak, then successfully started to ride the Microsoft SharePoint wave in 2003. Though KnowledgeLake remains a strong Microsoft partner, it does have its own independent information and automation platform and products.

Interestingly, the founders actually sold the company to PFU Ltd, an equity arm of Fujitsu, in 2014, then reacquired it in 2018 with backing from private equity firm Plymouth Growth Partners and First Bank. In 2019, the company acquired New York-based RPA start-up RatchetSoft. KnowledgeLake is headquartered in St. Louis, Missouri, and employs approximately 70 people, with annual revenue estimated at around $20 million.

The Technology

KnowledgeLake provides a wide range of enterprise content management (ECM) – aka content service – products. This report focuses on the January 2023 “Ontario” update to the KnowledgeLake Cloud platform. The platform provides a cloud-based, low-touch, document capture and processing platform as a service (PaaS) that leverages machine learning (ML) and artificial intelligence (AI), what Deep Analysis calls cognitive capture. We use the word platform, rather than product, as it consists of a pretty extensive set of functionalities that will be integrated, customized, and configured by a channel partner to meet specific needs.

The first thing to note about the KnowledgeLake Cloud is that it is cloud native. Interestingly, the firm has decided not to take the multi-tenant cloud route. Rather, every instance is separated and no customer data is co-mingled. Similarly, KnowledgeLake itself does not have any visibility into its customers’ data. The only thing shared in the KnowledgeLake Cloud is the cloud compute power. This caught our attention as it is not a common approach: multi-tenant architectures are typically much cheaper to run, hence that’s the standard model for SaaS firms such as Salesforce or NetSuite. For some, though certainly not all, potential KnowledgeLake customers, this added level of independence and security will be welcome. Furthermore, the KnowledgeLake Cloud uses a proprietary data gateway, a secure socket if you like, to enable customers to work in a hybrid fashion.

At the functionality level, the KnowledgeLake Cloud has end-to-end document processing capabilities from point of capture through classification, indexing, and workflows, to file and data storage. Embedded here are the usual components one would expect in the form of OCR/OMR at capture point and a repository to store documents, but where it gets interesting is in the availability of ML across the platform that can be trained to undertake basic, important manual tasks such as page and document separation to speed up the ingestion phase. A combination of rules and ML can also be configured and trained to tag, extract, and classify incoming documents.

At its heart, though, the ML and capture capabilities are designed to trigger workflows and ideally facilitate straight-through processing. And it is with big improvements to workflow functionality that the recent release of “Ontario” comes into play. Previously, buyers of KnowledgeLake with real workflow needs would likely have added on existing third-party process management tools such as K2 or Nintex. Though they can certainly still do that if they want to, KnowledgeLake now provides native workflow functionality as part of the platform. Previously it was limited to some RPA technology to eliminate repetitive tasks in the document processing lifecycle, but now it extends to workflow functionality that allows users to model, create, orchestrate, and manage end-to-end flows. Users can create granular permissions in the workflow process and implement conditional routing, looping, error handling, etc. In short, the new release has more than enough workflow functionality to meet most, if not all, typical document-centric requirements.

KnowledgeLake has also added improved electronic forms functionality to the platform, providing a user-friendly drag-and-drop style forms designer and builder. A single user interface (UI) allows users to view and manage all their activities (see Figure 1). The UI is underpinned by API orchestration, so associated and integrated third-party applications are managed here along with KnowledgeLake’s own functionality. The UI is well thought out, with many pre-built no-code/low-code options, widgets, and steps to drag and drop on the screen, to create workflows and new work activities in specific work environments. For example, this is where users access and work with ML via no- or low-code to create a new “fingerprint,” a unique identifier for a document type (for example, a specific supplier invoice). The ML capture capabilities have been designed to train on a single sample of, for example, a supplier’s invoice (or a small batch of more generic document types), essentially via a point-and-click action. Similarly, this is where users access all the classification, text analysis, or even document separation functions to drive intelligent process activities.

Perhaps equally notable in this release (beyond the workflow elements) is the increased provision of packaged vertical solutions: pre-configured applications to fast-track an organization to production. Currently available are industry-specific packaged solutions for regional banks and credit unions along with horizontal solutions for commercial loan and account onboarding. We expect more packaged applications to come later this year and into next. It’s an approach we strongly support; though no packaged application is going to be perfect out of the box, it is a means to fast-track deployment and reduce customer costs and complexity.

Ultimately, all cognitive capture products do much the same thing: capture documents and process them through a workflow. But they also typically require a lot of time to design, integrate, implement, and maintain, consisting as they do of multiple changing parts. However, KnowledgeLake Cloud has done a good job of bundling and building these parts and pieces into a single cloud platform that is exceptionally easy to use.