Why Your AI Strategy Needs Knowledge Centered Service (KCS)
Imagine you've developed a cutting-edge AI assistant for handling customer support requests. It’s powerful and capable of delivering human-like responses, but have you wondered how the AI will know what to say to your customers when asked specific questions about your business?
While conversational AI and large language models (LLMs) possess the capability to generate human-like answers, their knowledge is not all-encompassing. Dr. Andrew Ng Founder of DeepLearning.AI explains in the course Generative AI for Everyone (which I highly recommend if you’re new to this subject) that we can liken an LLM’s capabilities to those of a recent college graduate. To assess whether an AI could do something, ask yourself “Could a recent college graduate do that?” A recent college graduate would have very limited knowledge of your company at best and would require specialized training to handle specific information about your business.
One of the ways you could teach an AI about your business would be training it or using Retrieval Augmented Generation (RAG) on a large corpus of company data that many already have – your knowledge base. Documentation sources such as knowledge base articles are a treasure trove of unstructured data perfect for training an AI assistant. But how accurate is the information? Maintaining an up-to-date knowledge base requires continuous effort to align content with your product lifecycle and without regular updates, information will quickly become outdated. Without a coherent knowledge management strategy, even the most advanced AI assistant will provide inaccurate answers if it is not consistently maintained.
As we enter an age where AI and automation handles more customer support tasks than ever, knowledge management strategies must become an even more important priority in our businesses in order for these projects to reach their full potential. While AI technologies may be new, knowledge management (the creation, capture, structure, access, use, and improvement of information assets) has been well established for many years. Knowledge-Centered Service (KCS)®, the gold standard for knowledge management in customer support environments, encourages support teams to actively use and update their knowledge bases with every customer interaction. This methodology is used at Cisco by our Duo, Umbrella, and Splunk teams.
By following KCS practices, we create knowledge base content as a by-product of answering questions in a collaborative environment. When our support engineers handle a case, they use some of the following practices:
Capture the customer's context: Include the words and phrases they use to describe the issue, environment, troubleshooting steps, and resolution.
Search early, search often: Look up the customer's issue in the knowledge base to see if it exists. Someone may have asked that question before.
Reuse is review: If the article exists, use that article to resolve the issue and attach it to the case.
Flag it or fix it: If the KB article exists but it needs to be updated, correct the article to help the next person who needs to use it, or notify someone who can.
With KCS, new information is constantly being captured as we interact with customers. By consistently updating the knowledge base, AI assistants will be kept current, ensuring their responses are timely and accurate.
KCS also recommends having a content standard with a standardized structure for articles. Our teams who utilize KCS typically create knowledge articles in “One Question, One Answer” format, meaning that instead of lengthy guides, most of our articles focus on one single problem and solution. While we are still in testing, our findings so far seem to suggest that these Q&A-based articles have many benefits for use with AI - the shorter length consumes fewer tokens and reduces the need for section-based "chunking".
As a knowledge management leader, I believe an organization’s success in the age of AI will depend on its ability to effectively leverage its knowledge. Although it predates the advent of LLMs, strategies like KCS are ideal complements to an AI strategy and are essential for ensuring projects provide accurate responses over the long term.
KCS® is a service mark of the Consortium for Service Innovation™