Knowledge powered by artificial intelligence

Abstract

Generative Artificial Intelligence (GenAI) has revolutionized knowledge management, offering unprecedented capabilities for creating, proofing, summarizing, and evaluating documentation. This paper explores how AI, particularly large language models (LLMs), and Retrieval Augmented Generation (RAG) systems, can streamline the development of knowledge articles while addressing ethical concerns such as data ownership and bias. We examine practical applications, including real-time collaboration, multilingual support, personalized information retrieval, and automated knowledge forecasting. Additionally, we explore AI’s role in bridging legacy systems, reducing biases, and enhancing decision-making. Ultimately, AI extends beyond generating content, shaping a more efficient, inclusive, and innovative approach to knowledge management. This article is based upon a presentation given at the 2024 NISO Plus Conference that was held in Baltimore, MD, USA, February 13–14, 2024.

Five years ago, the idea of using AI to create documentation or knowledge articles was a novel concept, despite the technology’s existence. The emergence of Generative AI, ¹ propelled by the success and potential of ChatGPT, ² has ignited the imagination of many. This powerful technology has opened a world of possibilities, inspiring us to explore its diverse applications. Generative AI, in particular, has empowered us with a fantastic tool that can assist in developing, proofing, summarizing, and evaluating knowledge articles, paving the way for a future of knowledge creation that was once unimaginable.

The technology works with some assumptions—it will not always be perfect and will require human review for accuracy and relevancy. The technology requires a Large Language Model (LLM) ³ that can write or summarize the content, and this is usually where a lot of dialog begins about ethics and data. It is crucial to stress the importance of ethical considerations in AI usage, as it makes the audience feel responsible and mindful. Imagine a scenario where you are a managed service provider that provides technical support to businesses. You want to build up a knowledge base that helps your client’s self-serve support. A model trained on Apple Support documents will write excellent documentation since it has that data and can quickly write articles, but then the moral and ethical considerations of copyright and data ownership come into play, since, technically, the language and knowledge originated from an authoritative protected source.

Let us not get bogged down in that, but let us focus on a new idea that can bring immense value to you. Imagine a scenario that’s all too familiar—a library of information scattered across multiple systems. The goal is to streamline all this knowledge into one place and update articles. If as you read this you are nodding your head in agreement, this is for you. AI has the potential to revolutionize knowledge management, bringing unprecedented efficiency and organization to your work.

We can use a Large Language Model and point to an RAG ⁴ (Retrieval Augmented Generation) that contains all those articles. The best part of this is that we can do this entirely offline, too. There is no need to risk uploading data if you have private data about which you are concerned. Software tools such as LM Studio ⁵ and running an RAG off a laptop are entirely possible. What I have done with a lot of success is recording conversations (during a knowledge-sharing session), having an AI understand the conversation, and then building articles around what was discussed. You can quickly train these models to act as reviewers.

Let us model out the right kind of AI to take things further. We can rapidly develop documentation and knowledge articles efficiently if we train an AI model to ask questions as it reviews content and looks for knowledge gaps and critical points. Moreover, if we point the AI to a repository of information, we can have that model respond the same way. Taking one more giant leap forward, imagine loading your ticketing system data (as in the description of issue and resolution), determining what trends it can find that require an article, and having it write the article. Using AI to power through information and interact with it via a chat interface can elevate your knowledge, help you find what you are lacking, and build a more substantial knowledge base.

Another exciting area where AI can significantly enhance knowledge management is in the realm of real-time collaboration. As teams work on complex projects, knowledge sharing becomes crucial. Keeping documentation current can be daunting. AI can assist by analyzing project discussions, emails, and meeting transcripts to ensure that the knowledge base evolves in real-time. It can detect when certain concepts are mentioned repeatedly or when clarifications are made and automatically generate or update documentation to reflect these changes. This enables teams to stay on the same page without the administrative burden of manually updating documents after every meeting.

Additionally, AI can be leveraged to handle the growing complexity of data across industries. For example, in fields such as healthcare, where medical literature and research are updated constantly, it is nearly impossible for professionals to stay current with all the relevant studies and findings. Generative AI, combined with robust databases (RAGs), could comb through the latest research, summarize key points, and suggest how new findings could impact existing medical guidelines or protocols. This application saves healthcare professionals time and ensures that they are equipped with the most recent and relevant information when making decisions.

Another advantage of AI in knowledge management is its ability to personalize learning and information retrieval. AI-powered systems can learn from user behavior, preferences, and interaction patterns, tailoring content delivery to individual needs. In an educational setting, for instance, an AI could monitor students’ progress, identify areas where they struggle, and automatically suggest reading materials, practice questions, or additional resources. The same principles can be applied in a corporate environment where employees may require different training or support based upon their job functions or skill levels. Instead of sifting through massive knowledge repositories, users would be provided with the information that they need to succeed, improving their engagement and efficiency.

One exciting application is AI’s potential role in “knowledge forecasting.” Based on historical data, trends, and patterns, AI systems can predict what information will likely be needed and preemptively create or update content to meet those needs. For instance, in customer support, AI can analyze common inquiries and their resolutions, anticipating future questions arising as products evolve or new features are introduced. It can then proactively generate documentation or FAQs that address those potential issues, ensuring that users have the necessary support without waiting for problems to arise.

AI’s ability to interface with legacy systems also adds tremendous value to organizations that have vast amounts of data trapped in outdated formats. We can unlock this data and integrate it into modern knowledge systems by deploying AI. For instance, scanned PDFs, recorded audio, handwritten notes, or older proprietary formats can be digitized, understood, and incorporated into a dynamic knowledge base using AI-driven optical character recognition (OCR) ⁶ and natural language understanding. This ensures that valuable institutional knowledge is not lost but rather remains accessible and useful.

AI also offers the capability of automated multilingual support. In global organizations, documentation must often be created and maintained in multiple languages. This can be both time-consuming and prone to errors. AI models, trained in different languages, can simultaneously generate content in several languages while maintaining contextual accuracy. This ensures that consistent and reliable information is delivered across geographical regions without the need for multiple teams of translators. Furthermore, the integration of natural language processing (NLP) ⁷ allows AI to understand cultural nuances, thus creating content that is not only translated but that is also locally relevant and accessible.

Lastly, AI can play a significant role in reducing bias in knowledge creation. Human-created documentation often reflects unconscious biases or gaps in perspective. By using AI models trained to recognize and mitigate bias, organizations can ensure that their knowledge bases are more inclusive and equitable. This has profound implications in fields such as law, education, and healthcare, where the accuracy and fairness of information can directly impact individuals and communities.

AI is more than just generating content. AI’s role in knowledge management is transformative, extending beyond simple documentation creation. Its applications in real-time collaboration, personalized learning, multilingual support, knowledge forecasting, and bias reduction offer revolutionary possibilities for both individuals and organizations. By harnessing AI’s power, we are improving how we create and manage knowledge today and shaping a more efficient, inclusive, and forward-thinking future.

Statements and declarations