Key Takeaways
- Knowledge base taxonomy is the process of structuring your KB for easy searching and accurate information discovery.
- Good taxonomy requires smart governance and striking a balance between organization and tag-based searching.
- Human oversight and buy-in from Subject Matter Experts is a must.
- AI-powered systems can reduce the difficulty of KB management, but the AI must be carefully overseen.
What Is a Knowledge Base Taxonomy?
In this context, taxonomy means the structure of tags, categories, and other meta-data that makes it easier to locate information within your knowledge base. The better your KB taxonomy, the more quickly human workers and AI agents can locate critical information AND be able to rely on the information they find.
Good knowledge base structure can speed up activity and boost productivity across your entire organization. Every second an agent spends trying to find information, or writing new KB entries because they couldn’t find what they needed, is just wasted effort.
This can have further implications for large organizations that deal with extensive regulations or bureaucracy. Consider a biomedical organization that needs to adhere to different medical regulations across multiple states and countries. Their KB must contain accurate regulatory SOPs for each area and offer the correct SOP depending on the agent or customer’s location.
With poor taxonomy, there’s a real risk of the wrong procedures being followed, and legal actions as a result. Good, reliable taxonomy would guarantee the proper procedures are always found and followed.
Why Do Most Organizations Get KB Taxonomy Wrong?
In our experience, most organizations are still using outdated methods of creating their taxonomies. These methods, while common in the past, simply aren’t suited to modern situations where data in the KB is growing faster than people can stay on top of it.
In the past, most often taxonomy was determined by Subject Matter Experts (SMEs) manually reviewing, tagging, and categorizing articles. However, there are multiple problems with this:
- The process is slow and time-consuming.
- Data may be added faster than the SMEs can review.
- Multiple SMEs reviewing information may have different tagging strategies, leading to mismatched tags.
- Manual tagging often doesn’t take real-world search patterns into account, creating taxonomies that make sense on paper but fail to provide the information workers actually need.
- Past a certain KB size, fully manual taxonomy and validation may be genuinely impossible.
In addition, most companies creating KB taxonomies are only looking at the state of their Knowledge Base right now without preparing for future scaling. So they may end up trapped in a loop of having to review and re-tag their KB once a year or more, without getting ahead of the problem.
This results in hundreds of hours of manual effort, to minimal effect, and a workforce which is constantly slowed down by difficulty finding the information they need.
What Are the Core Components of a Knowledge Base Taxonomy?
Fundamentally, a knowledge base taxonomy is based around categorizing, tagging, and classifying knowledge to be as easily-discoverable as possible.
This includes:
- Topic Clustering: Associating related articles and documents based on their topic, to improve discoverability.
- Heirarchies: Breaking down topics into broad high-level categories which then break down into sub-category trees from there. This makes it easier for agents to ‘drill down’ when manually looking for information.
- Faceted Filtering: Treating clusters and heirarchy trees as filters, to focus on providing relevant information based on searches while filtering out irrelevant articles.
- Intelligent Tagging: Ensuring content tags are consistent across all articles, so that if an agent searches for something like “billing” it will bring all results regardless of other trees or topic clusters.
- Knowledge Intelligence: Combining KB taxonomy and smart AI systems to aid in proper categorization and tagging across large knowledge bases.
These techniques are complimentary, rather than exclusive. They’re all tools in the toolbox for creating a smarter KB that’s easier for agents and AI bots to search.
How Do You Build A Knowledge Base Taxonomy?
Your KB taxonomy will be specific to your needs and usage patterns within your organization. Here are some general guidelines for how to organize a knowledge base:
- Observe real-world usage. Start by gathering data on how your workforce (and AI, if applicable) are currently using the KB. What searches are run most often? What articles are most accessed? What improvements do users suggest? Your taxonomy should be developed “ground up” and based on actual usage.
- Bring in SMEs early. You need Subject Matter Experts who are fully committed to helping develop a better KB structure, as well as reviewing the articles themselves for accuracy.
- Develop the category structure. Your top-level categories should be clearly named, based on actual search patterns. Sub-categories should flow logically from high-level to detailed. Try to avoid more than 3-4 levels of sub-categories, or else discovery becomes difficult.
- Create a tagging governance policy. The tags associated with articles need to be standardized. Use a single clear-cut set of words and closely-associated synonyms, but be careful to avoid overlap or over-tagging the articles.
- Decide on content life cycle rules. How old is too old? The system should be tracking the life cycle of each piece of KB content, and flagging articles which are old enough to possibly be out of date.
- Integrate AI. If you’ll be using Knowledge Intelligence systems to assist with KB management, don’t get the AI involved until you’ve established the ground rules. And always have your human managers and SMEs in the loop, overseeing what it does.
- Leave room to grow. Your policies should be rigid, but not inflexible. Have enough procedures in place that new categories or search strategies can be deployed, if future usage calls for them.
How Is AI Search Changing the Rules of Knowledge Base Taxonomy?
AI-powered knowledge management can take a lot of the burden off human managers and SMEs, but it’s not a magic bullet.
The best aspect to using an LLM-based AI in KB management is that the AI’s vector search systems allow it to intelligently find links between different pieces of information. Perfect taxonomy isn’t as important when an AI is assisting with organization.
However, it’s still vital to have a clean structure and – most importantly – vet all the articles scrupulously for accuracy. AI is only as smart and accurate as the information it’s trained on. So your SMEs need to ensure the knowledge available is as accurate and non-contradictory as possible, before the AI starts training.
Better stucture, clean tags, and well-developed governance policies will all help keep the AI in line, as well.
What Are Common Taxonomy Mistakes and How Do You Avoid Them?
These are some of the issues we see most often, when we’re helping companies build better knowledge bases alongside AI agents to streamline knowledge retrieval.
- Over-tagging: Tags should be ‘right sized,’ covering the key topics in an article, but without tagging every concept. Too many tags can lead to agent searches with too many hits to quickly identify the proper article.
- Inconsistent Naming/Tagging Conventions: We strongly recommend creating an official list of approved terminology, to avoid use of synonyms. For example, always use “billing” OR “invoicing” in relevant articles but not both. This makes searching and AI discovery easier.
- Building Taxonomy in a Silo: Taxonomy policies shouldn’t be monolithic and entirely ‘top down.’ Real-world usage data, including feedback from users, is needed for best results.
- Failing to Revisit the Structure Over Time: Even a well-founded taxonomic base could prove unstable with enough KB growth over time. Be willing to occasionally rethink the entire structure to see if there are better approaches available.
Frequently Asked Questions
One is a subset of the other. Category structure is one part of a larger taxonomy. Categories organize the content, but don’t typically help discoverability. Taxonomy includes elements such as tagging and meta-data that improve discoverability for workers and AI agents.
As a general rule, somewhere between 3 and 7 tags is considered optimal, but don’t over-tag. The tags should be chosen to be MOST relevant to the information conveyed.
If you’re in a situation where an article covers multiple topics and would require more than 7 tags or so, consider splitting it into multiple linked articles instead.
Ideally, this is an ongoing process where SMEs work alongside AI agents to constantly review and revise the taxonomy day by day, with the goal of constantly pushing user search times down.
As a baseline, however, it should be reviewed at least once a year.
AI makes good knowledge management tagging and taxonomy even more important, because AI lacks human common sense and problem-solving abilities. All the AI agent knows is what’s in your KB articles and their metadata. So, the more accurate and reliable your taxonomy is, the more accurate and reliable your AI assistants will be.
In larger organizations, it’s typically best to have a dedicated management team overseeing the knowledge base. This should include at least one decision-maker who reports to the C-level, such as the CIO, as well as multiple SMEs and Information Architects handling the details.
However, individual departments should still have enough power to make KB updates based on their own needs, with periodic reviews of their work.
In Conclusion: Good Taxonomy Makes a Good Knowledge Base
As your knowledge base grows, so does the need for good taxonomy. Putting in time now to properly structure your KB, verify accuracy, and tag the contents will pay off with improved efficiency across your entire operation. Add AI to the mix, and you’ll have a knowledge base that’s ready to grow alongside your business for years to come.
KMS Lighthouse can make it happen! Our cutting-edge AI-powered knowledge management tools are already used by businesses around the world. Large or small, we can help you create a smarter, faster, better knowledge management system.
Click here to learn more, or contact us for a free demo.
