.png)
If you’ve been following along, you know we’re on a kick to demystify AI for education.
Our goal? Practical application over theoretical discussions.
One topic that’s been coming up—and a great use case for AI—is help center content. This includes both audio and video.
To get a better understanding of how AI can support the creation of help center content, we sat down with Neel Balar, Co-Founder of Clueso.
(Just a note: I’ll use help center and knowledge base interchangeably in this article.)
Why is AI so helpful for knowledge base content?
Knowledge base content is pretty straightforward: This is what the product does and here’s how you do it. Help content tends to follow a set of guidelines (Neel referenced Microsoft’s Writing Style Guide), making it more predictable and formulaic than other types of content. Content that’s clear and structured is a great use case for AI.
What’s the average time savings for a team using AI for knowledge base content creation?
Like with any form of delegation, there’s an initial upfront cost (time-wise) to train AI and build it into your workflow. This is true whether you’re using a purpose-built platform or leveraging a custom GPT for your content.
Neel shared that, on the low end, people are seeing a 60% time savings. But some Clueso customers are seeing a 98% time savings. Imagine a workflow 50x faster than your current one! 🤯
How are teams using AI to create knowledge base content?
Because knowledge base content is straightforward (this is how this works, here’s how to X in the product), it’s a great use case for AI.
That’s true whether you’re creating written help content or videos.
Depending on the AI tool you use, you can use AI to:
- Generate step-by-step instructions
- Create or edit screen recordings/tutorials
- Turn scripts into voiceovers
{{webinar-link}}
Is the AI just replacing certain steps in the usual process, or does it require a different mindset?
Neel encouraged us to think about AI as “another tool you can use to streamline your process and make it much, much faster.” So it may not replace certain steps in the process, so much as speed up those steps (especially time-intensive steps around content creation, editing, and review).
It may be helpful to think of AI as an extension of your team—one that helps you maximize the efficiency and productivity of your current resources. Like you would when delegating any task, be clear on what you expect and the directions you give. And just like you would with another team member, learn from each experience. If you don’t receive the output you were looking for, learn from that and change your prompt.
Where is AI streamlining vs. eliminating roles? What’s the role of a human in the process?
A big fear for many people is that AI is coming to take our jobs. Take heart, friend, because that’s not happening anytime soon.
Humans play a key role in using AI, from training the model to reviewing output. Specifically for help articles, someone on your team needs to be able to provide context to the AI, whether that’s screenshots or a screen recording within your product or through prompts. Once the AI has generated content, someone needs to vet the content for accuracy.
How do you decide what to share and what not with the AI system, based on security, privacy, and legal concerns?
This is an important question—and one every business is asking. Here are a few practices Neel shared:
- Remove personally identifiable information (PII). This protects the confidentiality of your customers.
- Read through the terms of service. Understand how your information is being used and stored. (If you’re using Open AI/ChatGPT, your information is being used to train the model. However, if you have the enterprise plan, your information is not stored or used to train.)
- Ask which AI models are used for a product. Most companies are tapping into larger models like Open AI or Google Gemini (formerly Bard). Your security team will likely dig into the models used, what’s accessed, etc. as part of their review.
- Be mindful of what’s in the content you share. Some information about your product is not confidential (especially when we’re talking help center content!). But be careful to remove any intellectual property/proprietary information from your AI prompts.
How do you ensure AI created content matches your brand voice style and lexicon?
In the same way you’d introduce technical writers and instructional designers to your brand voice guidelines and lexicon, you’ll share this information with AI. (Zapier has a great article on how to train a custom GPT. While this is specifically for Open AI/ChatGPT, you can apply these concepts to training other AI models.)
Neel told us that the key is to give AI all the guidelines that need to be followed—and they should be as descriptive as possible. Neel mentioned that examples are helpful to give the AI something to base its output on.
Here’s an example of how Neel would prompt the AI: “You are a technical writer working at company XYZ. Your job is to write the documentation on how each of these features in this product works. What I’m going to give you as an input is an overall broad feature overview. What you need to generate from it is a knowledge base article. Here are the guidelines you need to follow: Create it in a step-by-step format. Ensure that every step has a screenshot. Ensure that you’re starting with an action verb, keeping the voice in active voice. Make sure you’re following these brand voice guidelines.”
It’s helpful to keep in mind that the AI will do what it’s told, so be as specific as possible and avoid ambiguity.
{{demo-link}}
How much time do people spend copy editing AI-generated content?
Again, think of that employee context: As a new employee is learning your brand voice and lexicon, you’ll likely make more edits. But as that employee learns from the feedback, they create more quality content that requires less review.
In the same way, as AI learns what you expect and how to deliver it, you’ll spend less time reviewing and editing AI-generated content. Each time you review the output, you can modify your AI prompts to get better results. (You can see how Kristine Kukich did that in this blog, on example #15).
So the answer to this question varies, but both Neel and I have seen people accept roughly 60-70% of AI-generated content as-is, with the rest needing light modification or going back to re-prompt.
{{report-link}}
Embracing AI for Efficient Knowledge Base Creation
Integrating AI into your knowledge base content creation can revolutionize your workflow. By leveraging AI, you can streamline processes, significantly reduce the time spent on repetitive tasks, and maintain high-quality, consistent content.
As with any tool, the key is to use it thoughtfully and strategically, ensuring that it complements your team’s efforts. So, take the leap, experiment with AI, and watch as it transforms your content creation process for the better.
