Knowledge Management in the Age of AI

It’s safe to say that AI-powered tools have revolutionized searching, using, and creating information in all kinds of flexible, potent, and convenient ways.

Support organizations thrive on information, and have long strived to manage knowledge (create, maintain, organize, retrieve) formally and systematically in purpose-built knowledge bases. Have we now reached a point where we can jettison the very concept of the support knowledge base? Can we perhaps forget about knowledge management altogether? Can we redeploy humans to other tasks? Let’s find out.

Do we still need to create knowledge articles?

AI-powered search tools are so powerful that it’s tempting to just point them at case records, bug databases, internal and external discussion threads, and the like, and hope for the best. But the best may not be great.

• Extraneous information may confuse the search and return ambiguous or plain wrong recommendations.
• Existing records include obsolete information.
• Cases contain all kinds of information about customers, including code snippets and specific details on how they use the product, details that should not be exposed to the public. You can teach AI tools to obscure PII but there are many, more subtle confidential data points in your existing records.
• Both cases and bug databases contain confidential vendor information that you don’t want to expose to customers–and sometimes not even to internal users.

So the good ol’ knowledge base is still very much needed to feed the AI appetite. And remember we’re talking about maintainance, not just creation.

Can AI create knowledge articles?

Yes! AI tools are very good at creating knowledge articles from support cases and other artifacts, such as recorded conversations, webinars, etc. They can extract relevant information, discard trivia, and avoid article overlaps. They are particularly skilled at following formatting and branding requirements, unlike humans who are not particular good at it and often resent having to do it in the first place.

The strong recommendation is to institute some kind of human review of the AI drafts to avoid hallucinations and other tragic failures, but multiple vendors report that their AI tools create high-quality articles, the overwhelming majority of which are approved as-is by human reviewers.

What’s the future of KCS?

The venerable KCS (Knowledge-Centered Support) methodology, which guides support agents to search for articles, link articles to cases, and create articles starting during resolution to avoid duplicate effort is still completely valid. But it becomes much easier to use when paired with an AI-powered search tool and an AI-powered drafting tool.

KCS always billed itself as integrated into the case management process. AI tools allow it to achieve the integration seamlessly. I’m starting to see vendors stop talking about KCS completely, while using all its principles and leveraging AI for searching and drafting. It’s no longer a separate set of tasks, but instead a component of the overall case resolution process.

What’s the role of humans in knowledge management?

If AI is searching for and creating knowledge, what is left for humans to do? For starters, gathering and logging information–perhaps aided by other specialized AI-powered tools. Humans should also vet drafts of AI-generated knowledge articles, as mentioned above. And humans are needed to orchestrate the entire knowledge management process, and in particular to identify promising sources of knowledge outside the traditional boundaries of support or support cases.

On the other hand, AI tools can handle activities like searching, drafting articles, finding overlaps and duplicates, creating logical clusters of documents, even identifying knowledge gaps–and they can achieve much better results than humans.

What’s the longer-term future for knowledge management?

Support knowledge bases were invented because of weaknesses in the product documentation. Traditionally, product documentation is static and only changes to accompany product releases. And product documentation is often product-focused rather than use-case-focused, often because it’s slapped together on a tight schedule. Knowledge bases are meant to supplement the standard documentation with user-focused information that can be updated pretty much instantly.

As it becomes easier to manage knowledge, we may finally see a convergence of documentation and knowledge bases, creating a richer, more dynamic, unified source of information. Can’t wait!

Our Takeaways

• AI tools work much better against a properly-curated knowledge base.
• AI tools can draft excellent knowledge articles.
• The traditional KCS model is still completely relevant, but can be rebranded to better fit the current situation.
• Humans are the orchestrators, and can delegate much of the grunt work to AI tools.

How are you adapting your knowledge processes in the age of AI? Tell us in a comment.