Building a knowledge base for B2B SaaS support

Past tickets. Help center articles. Slack threads, Google Docs, slide decks. Most B2B SaaS support teams have plenty of knowledge available, but it’s scattered across the company’s tech stack. Agents struggle to find the information they need, and when they do come across it, they can’t always trust it to be up to date.

“Our tech stacks are absolutely insane. There isn’t a single source of truth. There isn’t just a knowledge base we can rely on.” — Josh Solomon, GM & VP of Revenue

The solution? Centralizing all that information in a knowledge base. 

But it’s not enough to write a few articles, tell the team they need to keep it updated somehow, and move on. The right decisions about architecture, content structure, governance, and tooling ensure a strong, long-lasting foundation before you write a word. And starting with a proof of concept will help you and your team be confident that you’re on the right path.

This post will walk you through best practices for B2B knowledge base setup and creating that solid foundation for ongoing knowledge base management.

‍

Why B2B SaaS knowledge bases are harder to build than they look

‍

Building a knowledge base for SaaS support in the B2B space is no small task. The products are technically complex, with frequent updates, and different customer segments can use them in wildly different ways. Knowledge base content needs to be as fresh and comprehensive as possible in order to keep up.

‍

And B2B customers demand more than in B2C. Whether they’re reading knowledge base articles themselves or getting support from a human or AI agent referring to those articles, they expect the information they get to be worth the money they’re paying.

‍

Don’t underestimate the thought, care, and time that should go into B2B knowledge base setup. But that investment can pay great dividends when it comes to customer support team efficiency and customer satisfaction.

“Support isn’t lacking knowledge. It’s lacking the ability to retrieve it.” — Josh Solomon, GM & VP of Revenue

Architecture: Help readers find the knowledge

Architectural decisions are foundational. They reverberate through all other aspects of your knowledge base and are critical to giving your team unified customer data for support, a key to providing great proactive support and scaling effectively. 

Think carefully about what’s right for your team, both now and in the future.

1. Audience

An internal knowledge base for support teams has different requirements than a customer-facing one—or one that serves both. While both audiences could find technical details relevant, your customers shouldn’t be able to read about internal support policies, so you’ll need to ensure the right articles are private. At the same time, there may be very different requirements for style, tone, and consistency depending on whether an article is for an internal or external audience.

Without clear guidelines around how to manage articles for the different audiences, mixing them together can cause confusion and complicate navigability and maintenance.

Once you’ve defined your audience, think about how they will likely find information: will they browse the knowledge base itself, use the search bar, or have it filtered via a chatbot or other AI feature? This answer plays a role in every other architecture decision.

2. Article types

Defining your audience can help you determine the types of articles they need. Here’s a common way to break these down:

• Conceptual: “What is X?”
• Procedural: “How do I X?”
• Troubleshooting: “X is broken!”
• Reference: Definitions, specifications, and other technical details

Diátaxis is another framework for knowledge article types. It’s a systematic approach to authoring help documentation, and it includes four key types of articles: tutorials, how-to guides, technical references, and explanations. 

Regardless of exactly which article types you choose, deciding and being consistent will help you organize your knowledge base and create templates to standardize future documentation.

3. Hierarchy

Hierarchy—or lack thereof—can affect both findability (finding information you already know you’re looking for) and discoverability (coming across information that you didn’t know ahead of time would be helpful). One page with a flat list of articles is certainly an easy place to start for a simple product, but it can seriously hinder navigability as complexity grows.

Adding hierarchy from the start scales much better, helping both humans and AI to find the right information for even a complex product. For most B2B SaaS knowledge bases, a shallow, two-level hierarchy of category and article can provide some basic structure when you’re getting started, with room to grow as your knowledge base becomes more mature. 

4. Taxonomy

The categories you choose—the taxonomy—make a difference, especially if customers are part of the audience. They think about tasks they want to complete and problems they want to solve, not feature names, so align your categories with the former rather than the latter. User research activities like card sorting can show how your audience naturally categorizes article topics.

‍

Content structure: Make the knowledge useful

Even the most carefully structured knowledge base won’t be useful if the content inside it is a mess. That’s why article structure is also so important. It usually includes three core components:

1. Article titles

Article titles should also incorporate language that the audience will actually use to search. “How to manage permissions” or “How to change what your team can see and edit” might make more sense than “User role hierarchy.” But be sure to include any relevant feature names or technical terms in the body of the article, so those who do search that way will still find the article.

2. Article body

In most cases, the article body should start with a short answer or summary, then provide the full explanation. Readers want the key information right upfront, with the option to dig deeper if they have the time and inclination—so don’t bury the lede. Many AI agents also weigh earlier information more heavily.

A few more notes on the body:

• If an article is only relevant to a subset of the audience—say, your Enterprise tier—mention that in the first sentence. That way, Pro-tier customers aren’t left wondering why they can’t find the feature. 
• The length of the article as a whole depends on the information that needs to be conveyed. A 400-word article that provides a single clear path can be more helpful, and less confusing, than an 800-word one with a few optional paths. Intrepid customers will discover alternatives on their own.
• If you do find your articles getting long, consider breaking them up into multiple shorter articles with interlinking to help readers navigate through them.

Your article body may also include other types of content, like product screenshots, embedded videos, or other relevant resources.

3. Metadata

Metadata—not the content itself, but the data about that content—provides context that’s essential for both ongoing governance and successful AI retrieval.

For each article, consider including the following information, depending on what’s possible in your knowledge base tool:

• The owner
• The date it was last reviewed
• Who it applies to (for example, Pro vs. Enterprise tier, or end users vs. developers)
• Related articles

Knowledge base governance and maintenance

The support knowledge base structure you and your team put time and thought into, the articles you carefully craft—without proper knowledge base management practices in place, all of that can quickly go to waste as bloat and stale, incorrect content take over. The details of maintenance are for another article, but they depend on a solid plan for governance from the start.

“Without accurate intake, search can’t learn. Patterns can’t emerge. Knowledge can’t evolve. The lifecycle just resets and repeats every morning.” — Tina Grubisa, Head of Value Consulting

Set clear ownership

Maintenance won’t happen if no one is explicitly responsible for it. On the other hand, assigning everything to a single knowledge manager can be overwhelming and most likely isn’t efficient. If you assign article owners by product area or team, the people closest to each feature will know exactly what needs updating. And a knowledge manager can still make sure those experts stay on top of their articles.

Develop a review cadence

The exact review cadence depends on things like how often your engineering team pushes out new features, but regardless, your top viewed articles should be reviewed frequently, perhaps as often as monthly. And whenever a new or updated feature is slated for release, up-to-date documentation should be part of the “definition of done,” not an afterthought.

Historically, regular knowledge base audits ensured that nothing falls through the cracks, like outdated info or broken links. That’s changing with the implementation of AI across customer support—more on this below. 

Track key metrics

Metrics can give an idea of knowledge base performance and help guide prioritization of which content should be added, reviewed, and removed. Here are a few that can help you understand your knowledge base’s effectiveness and health:

• Number of searches with no results (and what the search terms were)
• The rate of knowledge base visitors relative to support tickets submitted, also known as self-service score
• Article views or agent usage rate
• Average last-reviewed age

How AI changes what a knowledge base can do

Just a few years ago, a knowledge base was essentially a passive repository of information. Agents and customers had to first find their way there and then figure out how to find the answers they needed. 

Today, a knowledge base is so much more—it’s a key source of knowledge for every AI-powered function across your support stack, including self-service chatbots, agent assist, and ticket routing.

That means that knowledge base quality is even more important. Out-of-date content spreads beyond the confines of the knowledge base, and a disorganized structure makes AI retrieval difficult. Luckily, the inverse is also true—a well-structured, well-maintained knowledge base makes every AI function more effective, increasing efficiency and customer satisfaction.

Fortunately, AI can help with creating and maintaining your knowledge base. 

AI-powered knowledge management gives you more than a static help center reviewed once per quarter—your KB becomes a living, continually evolving asset. A customer support AI platform can monitor your customer interactions—support tickets, Slack channels, and call transcripts—to surface recurring customer questions and knowledge gaps. Once those are identified, AI can also automatically create new knowledge base articles for your team, creating a great feedback loop that continually gets better.

Even if a full customer support AI platform is an OKR for a future quarter, keep it in mind while you’re in this knowledge base building phase, including when you choose a tool. Ensuring consistent article types, clear titles, useful metadata, and up-to-date content now will make AI implementation much faster and easier when you get to it.

Of course, all of these points will also provide your current human readers with the best knowledge base experience possible.

Tooling: Choose a home for your knowledge

First determine what you want and need for your knowledge base, then find the tool that best fits those requirements. The tool should adapt to your team, and not the other way around.

The main options here are:

• Products embedded in your legacy helpdesk solution (like Zendesk Knowledge or Freshdesk Knowledge Base): With a familiar interface and simple features, these can be the lightest lift for teams just getting started but might not be enough for anything beyond basic self-service.
• Purpose-built tools (like Guru, Notion, or Confluence): These offer more options for structure, more robust governance, and better UX. But single-purpose tools also make the toolbox even more crowded, adding one more integration to maintain.
• Unified AI platforms (like Mosaic AI): These house your knowledge base under the same roof as your agent assist and self-service AI. They do a lot more than just knowledge management, so you’ll need to prioritize the problems you want to solve during implementation. But they also provide the greatest benefits, as you see the articles starting to improve the AI-powered functions—and vice versa.

Before you make a final choice, ask yourself: Who will use this knowledge? How will they access it? How will they use AI? How does this fit into our broader support tech stack and strategy?

These key points and everything else discussed in this article will help you hone in on the right knowledge base tool.

Getting started with your first knowledge base

You’ve made your key decisions and you have your tool of choice. Now you want to get all that useful information out into the world, and it can be tempting to dive headfirst into writing. But you don’t want to spend weeks manually writing 200 articles only to find out that they’re not giving customers the information they need. 

To determine what to focus on first, start by identifying support knowledge gaps based on your top ticket categories over the last 90 days. If you had articles covering those topics, how many of those tickets could have been resolved faster—or wouldn’t have been sent at all?

If you’re using an AI platform, this is really easy. Connect the AI to your support helpdesk, and let it answer those questions by analyzing ticket data. Based on how your team answered past tickets, you can then have the AI generate the articles. 

While AI supercharges this effort, I recommend getting your support agents involved too. They know which questions keep coming up and which existing answers aren’t up to date. They’re closer to your customers than anyone else at your company, and giving them some input as you build your knowledge base will help build trust and buy-in.

Once your first batch of articles are published and in front of your customers, watch your metrics and other data to see what’s working, what’s not, and what to prioritize next.

Set your knowledge base up for success

Writing  is essential to building a knowledge base for B2B SaaS support, but it’s not the only piece. You also have to consider architecture, governance, and, increasingly, AI infrastructure and implementation.

Think through the overall structure, give every article an owner, and start small to make sure you’re going in the right direction. And if you’re on the hunt for the right tool, request a Mosaic AI demo to see how it can help you build an incredible B2B knowledge base, improve self-service, and scale your support team more effectively across your customer base. 

FAQs

1. What should a B2B SaaS knowledge base include?

A B2B SaaS knowledge base should have clear categories and four core article types: conceptual (“What is X?”), procedural (“How do I X?”), troubleshooting (“X is broken!”), and reference (definitions, specifications, and other technical details). It can be for support agents, for customers, or both. Regardless, it should be clear which articles apply to which audience, pricing tier, and use case.

‍

2. How do you structure a knowledge base for B2B support?

Both overly flat and overly complex structures can be difficult to navigate. Start with a shallow, two-level hierarchy of categories and articles. Category names should be based on customers’ “jobs to be done” and the language they use, not product feature names. Clearly delineate your article types, and start each article with a short answer or summary before providing more details. Metadata like article owner, last reviewed date, who the article applies to, and related articles supports governance and AI retrieval.

‍

3. What’s the difference between a knowledge base and FAQs?

FAQs are direct answers to common questions, while a knowledge base provides a fuller picture: detailed explanations, step-by-step procedures, and more. Both can be useful for B2B support self-service, but FAQs should clearly be a shallower depth than the knowledge base, not duplicate it. 

A FAQ entry can give a quick “Yes, this is possible” answer and link to a knowledge base article with step-by-step instructions.

‍

4. How do you keep a knowledge base up to date?

Clear ownership and ongoing improvements are essential aspects of knowledge management. Assign article owners by product area or subject matter, and make sure knowledge base updates are part of each product release’s “definition of done.” Use customer support AI to identify knowledge gaps and create new content, and track metrics like self-service score and article views to understand which content needs improving. 

‍

5. How does a knowledge base connect to AI-powered support tools?

AI can help build your knowledge base, which then becomes a key training source for other AI-powered support functions like self-service, chatbots, agent assist tools, and more. An effective knowledge base improves your customer service on multiple fronts—it increases self-service and enables human agents to respond to customers faster and more thoroughly.

‍