Maintaining a deep catalog of knowledge articles to inform end users is an essential — yet often tedious — requirement for IT teams. But generative AI is offering a big assist. 
By using AI in knowledge bases, service desk agents who used to spend two or three hours creating each knowledge article from scratch can create drafts in minutes. And not only is today’s AI faster, but it also lets you personalize the knowledge you share — including translating your articles for a global workforce.   

Now, you can tailor the knowledge you share based on context — on the persona the system is interacting with — to serve unique needs and preferences. 

Why is this more important than ever? Because according to Ivanti research, more than three in four IT professionals say work stress is affecting their physical or mental health, and 68% say they feel burned out by their work. These mental stressors can drive up turnover and put pressure on productivity.  

Removing some of the manual tasks previously required to manage your knowledge materials relieves a great burden from IT support personnel — while also improving the overall employee experience. A purpose-built knowledge management AI can automatically update an organization's knowledge base using existing helpdesk data, then turn the knowledge base into a helpful copilot for IT analysts, arming them with relevant information to handle higher-order tasks.

Multimodal models change the game

Early large language models (LLMs) were focused more on processing, meaning they were confined to solving linguistic tasks. New multimodal models can process much more — including audio and video.

Gen AI can keenly analyze user behavior to provide personalized support and smart suggestions, for instance:

  • Multiple responses can be generated from a single document to personalize messages for diverse audiences and channels.  
  • By training gen AI with data from use cases across your company, it can then apply LLM to interpret queries from your end users, establish the best knowledge base match and retrieve the most relevant information.

Users expect everything they need to be available at their fingertips on their channel of choice. For instance, users might rely on a mobile communication application like Teams or Slack. However your users communicate, they expect the information to be presented on their channel of choice in a way they can quickly absorb.

Say an IT team must troubleshoot an issue flagged by a digital assistant, a chatbot or a copilot. AI can support an intelligent ticket routing solution so the IT helpdesk can respond faster. And, when the ticket progresses to the next level, the AI solution can help automatically triage the ticket for resolution.

The three R’s guide AI for knowledge management

Guiding the lifecycle management of knowledge articles using AI can be achieved easily by following these three principles, which I call the three R’s: relevance, revision, and retirement. 

  • Relevance: AI helps serve articles to users based on preference and usage and can recommend articles based on what is relevant for the user.
  • Revision: AI can provide information about articles that are not being used much but could become more useful if the content is tweaked — and then be used to revise those articles.
  • Retirement: AI can be used to indicate which articles simply need to be removed from circulation.

Five steps to writing knowledge articles with AI

Now that you know how AI can help your IT service desk automate the creation of knowledge articles, here's how I would approach creating a workflow to implement that capability.

  1. Identify specific challenges and opportunities you want to address. Focus on your key pain points. You can start with a data-gathering brainstorming session that involves multiple departments in your organization all looking at your existing system.
  2. Define clear goals and expectations for integrating AI into your knowledge management, such as increasing adoption of knowledge articles and improving productivity. Are you, for instance, trying to increase the number of knowledge articles in your database, or refresh a set number of knowledge articles while removing a specific number of outdated articles?
  3. Choose the right AI technology for your knowledge system. For instance, you might introduce natural language processing, the semantic search vector DB, machine learning for predictive analytics, automatic recommendations or a chatbot that integrates well with your knowledge article solution.
  4. Data security and access control. Make sure the data you provide is the most relevant and the most accurate and is presented in a well-structured manner to the AI. At the same time — since it is gathering data from different sources — it is vital to protect sensitive information and ensure you are complying with data privacy and other regulations. Ensure you put proper entitlement in place so users access only the information that they are allowed to view.
  5. Finally, it comes down to execution. Start small by looking to solve a minimum sellable use case — a pressing pain point. Then, try that with your initial set of users in a phased rollout. Try to fail fast. If something doesn’t work, iterate through it. Gradually, you will be able to expand your user base to all parts of the business, then apply continuous monitoring.

Once you’re running comfortably, that’s when you’ll start seeking feedback from users. As you start seeing improvements in user experience and productivity, you’ll know you’re on the right track. Just keep iterating through that process, and you will see your efficiency take off.

Read more in our reportHow AI is redefining IT service desk automation