Why Your Agents Hate Your Knowledge Base (And How to Fix It)
One of the biggest problems I’ve faced as a knowledge manager in customer service contact centers falls under the umbrella of agent adoption. I’ve run projects where we replaced an old, clunky knowledge base—something built in SharePoint or OneNote (yikes)—with a brand-new system in Salesforce, complete with a much-improved search algorithm. And yet, even after the upgrade, I’d still hear the same complaint from agents:
“I can’t find anything.”
It’s a frustratingly common problem. Even if you have a good system, the KB often becomes a source of irritation rather than a tool for success. This isn’t just about one thing; it’s a failure on multiple fronts.
First, there’s the search itself. Even with powerful algorithms and suggested content, older keyword-based search is flawed. We used to spend so much effort maintaining lists of keywords for each article, but agents would still get a dozen results with the same tags. They’d have to click into each one, read a bit, and back out, all while a customer is waiting on the line. That wasted time is a direct hit to your metrics.
Second, the information itself isn’t reliable. Agents search and find multiple articles on the same topic with slightly different, contradictory takes. They use one, get burned in a QA evaluation or anger a customer, and immediately lose trust in the entire system. Content health is crucial; if agents can’t rely on what they find, the tool is useless.
Finally, the design and structure of the content is often a major pain point. Agents aren’t knowledge managers; they’re problem-solvers under immense pressure. They need information structured logically and intuitively.
I remember working at one company where we had a very detailed policy for handling certain issues. The subject matter experts were adamant that it needed to be one big document with everything included. I really had to fight the battle to break it up into digestible chunks. When we finally did, the adoption and user ratings for that content went way up. It became understandable.
These failures have real consequences. They create longer handle times, lead to inconsistent answers across channels, and erode customer confidence. Most importantly, they burn out your agents. When agents feel uncertain and get beat up every day because their tools are failing them, they’re not going to stick around. All the money you invested in training and retaining them is wasted.
So, how do we fix it?
Improving discoverability is paramount. Modern, AI-powered search helps a lot because it can better predict the agent’s intent. An agent co-pilot that listens to the conversation and suggests the right content in the moment is a game-changer.
But the most overlooked solution is to engage your agents in knowledge creation. They can’t just be consumers of knowledge. Methodologies like Knowledge-Centered Service (KCS), which the Consortium for Service Innovation has been developing for nearly 30 years, are built on this idea. In KCS, knowledge is created as a by-product of solving problems. Agents are actively involved in creating new articles and flagging what’s outdated or conflicting.
To make this work, you need to build great feedback mechanisms and incentivize their use. Maybe you create a program that rewards agents whose feedback results in a change to an article. Give them the tools to create and share knowledge when they come across something new. New AI tools can even help, taking case notes and conversations and drafting an article for an agent to validate.
The knowledge they’re sharing and updating must be structured correctly. This isn't just about making it look neat; it's about making it usable. Think like an agent and organize information in the natural order of a call: greeting, authentication, troubleshooting steps, and so on.
Use formatting that makes information scannable—bullet points, highlighted scripting, and clear headings instead of long, dense blocks of text. You also need different article formats for different needs. A brand-new employee needs a verbose, step-by-step guide, while a veteran agent just needs a bare-bones refresher. (Exposing content based on roles is a powerful way to help improve discoverability, too.)
These days, you also have to structure content for AI. This means using standardized templates, like the "Problem-Solution" format, and clearly labeling sections so that bots can digest the information correctly. A good practice is to include a detailed procedure for the main body of the article but also add a list of FAQs at the bottom. This gives both agents and AI the best of both worlds: a comprehensive guide and quick, specific answers.
The bad news? This sounds like a lot of work.
It is.
The good news: AI tools can help with all of it. It’s not here to replace your agents, it’s here to help them–and by extension, your customer service team–work smarter and faster. By engaging your users in the conversation, you amplify acceptance and adoption.
Ultimately, this is about more than just technology. It’s an organizational transformation. The Boston Consulting Group calls it the 10/20/70 rule: 10 percent of thought toward the algorithm, 30 percent toward technology and 70 percent on organizational transformation.
Starting with a holistic approach and focusing on the human element is the key.
