By Subrata Chakrabarti
Every enterprise shares a common challenge: they need to be more agile and responsive to customer needs and market changes, but the pace of change makes it difficult to assemble all the right information in one central location. To be effective, their employees need access to intelligence that lets them answer the tough questions and win over customer confidence. Whether you’re an insurance agent investigating claim accuracy or a manufacturing shop-floor employee searching for a missing part of a machine, the ability to quickly locate and act on the right information will make all the difference.
Enterprises have amassed tremendous volumes of valuable data that could help them make better decisions and improve productivity, but it’s held captive in unstructured text documents such as claim reports, support tickets, manuals for parts and machinery, research documents, quality inspection reports, emails, etc. To date, AI offerings have had only marginal success in attacking this problem and solutions have been time-consuming and expensive for enterprises to create. Typically, their efforts failed for one of two reasons. The teams either pursued a strategy that was too broad—attempting to uncover all unstructured data to solve every potential use case—or they went too deep, pursuing a use case that required too much historical data, took too long time to solve, and could not scale.
But that was yesterday’s AI.
In IDC’s recent PlanScape, AI-Powered Knowledge Bases, IDC analyst Hayley Sutherland argues that there’s a new way to solve this problem, identifying a proven formula to unlock this intelligence and quickly fuel knowledge worker productivity, lower costs, and reduce business risk. She also provides guidance about how to select the right starting place, focusing on use cases in which the intelligence derived from unstructured text can provide a significant competitive advantage and ROI.
In the report, IDC profiles the Kyndi products and best practices that helped Siemens build its own AI-powered knowledge base, requiring just 3 months from concept inception to results evaluation. During their comprehensive PoC, they transformed a massive, unwieldy corporate wiki into an automated, NLP-enabled chatbot. Sutherland provides enterprises with the following four insights:
- What an AI-powered knowledge base is and the types of use cases it enables.
Today’s knowledge workers waste an average of nearly 10 hours per week hunting through multiple content repositories and searching for, but not finding, the right information. “As the productivity-enhancing unstructured content they need continues to grow at more than 40% CAGR per year  there is a clear need for a way to capture, organize, and analyze the knowledge so that it can be delivered to the right worker at the right time.”
The use case enterprises select plays a pivotal role. To be successful, enterprises need to stay laser-focused on maintaining the balance between managing complexity and the magnitude of the value they can uncover. The solution design must focus on the needs of the specific knowledge workers who will use the intelligence. Ease of use and simplicity of access are the key design principles. The ability to scale seamlessly while accommodating other use cases must be built into the platform foundation. Such is the vision of an AI-powered knowledge base.
Sutherland believes that once the knowledge base has been created, it can become the cornerstone to support a variety of enterprise functions. “By bringing together disparate types of organizational content, from customer service records to shipping information to maintenance reports, it provides the foundation for use cases from intelligent enterprise search to question-answering chatbots to IT ticket support and even helps these kinds of systems work together.”
- Why acquiring this capability is more important now than ever.
According to Sutherland, it’s never been more essential for enterprises to empower their employees with an AI-powered knowledge base. “With most knowledge workers performing their jobs remotely these days, the ability to self-serve and quickly find actionable, relevant information is more critical than ever.”
- Who will benefit from an AI-powered knowledge base.
Sutherland makes the case that line-of-business managers (LoBs) and a wide range of IT roles will all ultimately benefit from a powerful, unified knowledge base that can access and cultivate unstructured text and put it to work. LoBs are on point to digitally transform processes to accelerate sound decision-making and lower costs.
CIOs, CTOs, and information and knowledge architects all have a stake in supporting LoB initiatives. They can build their LoB constituents’ confidence and trust by successfully deploying and proving out one use case at a time—establishing a track record of delivering measurable improvements in productivity, cost savings, and reduced risk. As this process matures, the ability to extract intelligence from unstructured text can become a core competency for a broad range of enterprise organizations over time.
- How to build an AI-powered knowledge base.
“The first step in building an AI-powered knowledge base is to create a single source and unified view of the truth.” Sutherland documents the PoC process and best practices Siemens used as it leveraged Kyndi machine learning and natural language processing capabilities to build an AI-powered knowledge base that supported the Siemens’ Competence Center for internal and external financial reporting.
Siemens did an excellent job finding the right balance between managing complexity and delivering outstanding business value. They focused their PoC on a specific, business-critical use case that had the potential to scale enterprise-wide. They also identified a relevant measurement methodology that could quantify productivity gains and they obtained end-user input throughout the process.
During the PoC, Siemens deployed Kyndi solutions to create a comprehensive semantic knowledge base. Siemens then added Kyndi’s robust query processing engine to answer any natural language (NL) question, including cognitive search that went beyond keywords to include concepts and identify associations among them. Even when the discovery started with one word, phrase, or an incomplete question, the journey ended with a broader context and understanding.
Siemens set up a robust, iterative testing strategy that assessed both out of the box performance and performance with training, evaluating Kyndi’s performance consistency on a variety of data within the wiki. The testing phase took only two weeks and confirmed that Kyndi’s ability to perform a more sophisticated conceptual search resulted in a much higher success rate for financial analysts trying to locate information. The PoC outcome: the foundation for a chatbot was created that could significantly improve user productivity and lower support costs for more than 900 sub-organizations within the worldwide Siemens corporation.
Key Takeaways for Enterprise IT and Line of Business Managers
The IDC PlanScape AI-Powered Knowledge Bases report provides exceptional guidance for enterprises seeking a proven way to unlock the intelligence in their unstructured text and debunks the myth that an evaluation process must be long and expensive, with uncertain outcomes.
Join IDC and Kyndi for More In Depth Analysis
If you want to hear more about the value of AI-powered knowledge bases or the successful PoC process Siemens deployed, please join IDC analyst, Hayley Sutherland and Kyndi for our upcoming webinar — date and time to be announced shortly. We will focus on how organizations can leverage the same solutions and implement the same PoC process to dramatically improve productivity and decision-making with unprecedented speed.
 IDC Survey: AI-Enabled Enterprise Search 2019 Trends
 IDC Worldwide Global DataSphere Forecast, 2020–2024: The COVID-19 Data Bump and the Future of Data Growth, IDC #US44797920, April 2020