Implementing trustworthy and reliable KM generative AI

The power of generative AI (genAI) is tied to its endless applicability; enterprises in every industry are imagining ways where genAI can enhance the ways their business runs. In the realm of KM, its prospects are no different. While genAI can promise a world of possibilities for transforming knowledge management in the enterprise, it does not come without its challenges.

Jeff Evernham, VP of strategy and solutions at Sinequa, and Ryan Welsh, founder and CEO of Kyndi, joined KMWorld’s webinar,  Applying Generative Language Models to Enterprise Knowledge Management, to explore strategies and best practices for implementing genAI with accuracy and reliability at the forefront of its execution.

According to Evernham, knowledge management most often starts with search. And with the proliferation of genAI and the large language models (LLMs) that power them, communication between humans and computers has gotten a lot simpler.

While many fear that genAI is “replacing” search, Evernham argued that it is simply complementing it. Meaning, LLMs enable true semantic search, inviting more accurate results and better relevance—which is, at the end of the day, only amplifying the efficacy of what search is already designed to do.

At the same time, Evernham acknowledged that though genAI like ChatGPT is amazing, it doesn’t truly know what it produces. LLMs simply generate, and do not necessarily retrieve information from a reliable source.

With Sinequa, organizations can leverage the power of both search and genAI, where any genAI tool can be used to interface with the knowledge that already courses through your enterprise. This allows you to “Converse with Your Content,” where the LLM acts more similarly to a highly educated and trusted employee rather than a search engine, armed with configurable prompts, configurable UI, configurable conversations, and more.

Welsh continued the conversation by identifying the challenges of knowledge management, including:

• Difficult to measure ROI
• Outdated, conflicting, and siloed information
• Managing growing volume of information
• No easy way to add new or domain-specific content to overall knowledge base
• Lack of effective tools to understand how well you are managing your knowledge management

Furthermore, he identified the challenges of genAI:

• Inaccurate information (“hallucinations”)
• Concerns over security of proprietary information
• Lack of explainability
• Lack of knowledge of recent and domain-specific information
• Cost

According to Welsh, the solution lies with the Kyndi Generative AI Answer Engine, an enterprise-grade, end-to-end solution—building off  the power of genAI and LLMs—that provides users with direct, contextually relevant, and trustworthy answers instantly for enhanced decision-support, productivity, and customer experience.

Answer engines solve the challenges associated with genAI, Welsh explained. Kyndi’s solution is constrained to enterprise content, unlike a public genAI, allowing it to generate information based solely on the resources that an enterprise provides.

Kyndi’s Answer Engine further comes with:

• Enterprise-grade security measures
• Listed sources for each generative summary
• Simple system tuning for incorporating new and domain-specific information
• AI analytics for content insight and creation
• Auto-updated information when underlying content changes
• Auto-generation of topics and questions
• Lower query cost

For an in-depth discussion of reliable genAI strategies and solutions, featuring demos, use cases, and more, you can view an archived version of the webinar here.