Smart research, better docs: Leveraging AI in technical documentation

If you're like most documentarians, you've probably spent hours diving deep into technical research, only to emerge with way more information than your readers need. Sound familiar?

While good research translates into good documentation, it's all too easy to get bogged down with the details. Margaret Eker, a tech writer at Magento known for her research superpowers, tackled this exact problem.

Six years ago, she shared some strategies for boosting research skills on The Not-Boring Tech Writer podcast. With generative AI in our toolbox, this advice has become even more powerful.

Let's explore some use cases for leveraging AI to enhance your technical documentation:

Streamlining information discovery

As Eker points out, feeling overwhelmed often happens when you're just trying to grasp the basics—terminology, core concepts, and key players. Today's constant stream of updates only amplifies this challenge.

Documentarians can get stuck digging through:

• Tickets and user stories
• Technical explanations and product demos
• Interview responses from subject matter experts (SMEs)
• Countless articles, websites, and podcasts

Thankfully, AI can help with:

• Summarizing source material: Lex AI, a writing assistant, helps me distill key points from SME conversations.
• Identifying key concepts across multiple sources: Perplexity, an AI-powered search engine, analyzes multiple sources to give me comprehensive summaries.
• Organizing research findings: When writing about surveillance tech for Crikey, Lex spotted gaps in my research and showed me exactly where I needed more examples and quotes.

My various AI research and writing partners helped me figure out where to dig deeper!

Mastering domain knowledge

Once you've mastered the basics in a given domain, you need to understand the business details behind them. As Eker notes, business concepts often hide behind deceptively simple field names in software.

For example, a developer may combine IDs from last names and emails and call it "customer." But the business might include account numbers in their "customer" concept—the same terminology in a different context.

To get a handle on domain knowledge, technical writers can turn to the experts. Here is how generative AI helps me bridge the gap:

• Identifying dependencies: While researching knowledge graph systems, Lex AI caught my confusion between "property graphs" and "knowledge graphs." It explained the difference by using a family photo album. Property graphs just tag relationships, while knowledge graphs understand how relationships work—like how families gain in-laws through marriage.
• Finding knowledge gaps:  In writing about data governance practices, AI spotted where I needed more detail. Its suggestions prompted me to expand my description with a real-life data quality issue. This addition helped me clarify how data governance relates to AI governance.
• Checking assumptions: When explaining data as a product (DaaP), I explain that it emerged from data mesh. I had Lex check my assumptions about how it did so and discovered that I'd skipped defining "domains" adequately—a key concept readers need to understand. So, I further developed my reasoning and content to meet this need.

Partner with AI to spot inaccuracies and gaps before your readers do. Ask it to check your assumptions and highlight where you need stronger examples.

Creating user-focused documentation

Translating domain knowledge for multiple audiences can be a big challenge! Each audience—whether developers, support teams, or non-technical users—requires a different approach.

I find generative AI extremely helpful in understanding different user perspectives and tailoring my writing accordingly. Before AI, I spent significant time searching the web for examples of content style and technical accuracy for each audience type.

Now, with my partner, researching and understanding user needs has become much easier. AI can assist documentarians in:

• Adapting technical content for different levels: I write an opinion piece on surveillance technologies for a public news source (Crikey). When trying to explain surveillance to a general audience, my initial draft was too technical: "Law enforcement agencies obtain data from surveillance technologies, equipment, or information systems to ensure community safety.“ Using generative AI, I simplified my explanation of surveillance technologies to a “means to ensure community safety.”
• Identifying audience-specific needs: When writing about healthcare workers wearing body-worn cameras (BWCs), I researched healthcare user concerns. I turned to the Perplexity AI search engine, which highlighted the need for security and patient confidentiality. This helped me focus the article on the critical decision points and benefits for BWCs rather than their technical specifications.
• Generating examples: When writing a different article for KnowledgeOwl about documentarians and AI being partners in creating content, AI assisted me by showing how it adapts content for different audiences. I provided AI with DATAVERSITY and KnowledgeOwl articles as examples of style and tone. Then I asked it to change content that was written in the DATAVERSITY style to more closely match the style and tone of KnowledgeOwl's content.

Through these examples, you can see how AI can help identify where readers might need more context or explanation.

Maintaining document quality

Speaking of accuracy, let's talk quality checks—we all likely know their importance! Eker gives good advice when she said that good research helps us enforce precision and reliability.

Here's how I use AI to make my documents better for my readers:

• Consistency checking: DATAVERSITY had specific conventions in using and capitalizing terms. For example, according to its style guide, “Data Quality” needed to have complete capitalization, even in the middle of a sentence. I input the guide as a prompt and instructed Lex AI to review my DATAVERSITY articles to ensure they consistently adhered to DATAVERSITY's style guide.
• Validating technical accuracy: I wrote a DATAVERSITY article about industry experts' insights into the Age of AI. I shared the presentation transcript (the basis for my article) with Lex AI, and it helped identify where my content structure needed stronger connections to the experts' key insights.
• Aligning style and terminology: I wanted my data governance article to match DATAVERSITY's tone. To achieve this, I found three good examples of DATAVERSITY's tone and asked AI to analyze and compare them to my data governance article.

Quality matters—especially when you're getting ready to share your docs with readers. By using AI to help with quality checks, you can spend less time digging through source materials and more time making your content better.

Conclusion

Research doesn't have to be a time sink. By combining Eker's proven strategies with AI tools, we can make our research process both efficient and accurate.

I've found that AI works best when using it to summarize information, check domain knowledge, tailor content to different users, and maintain document quality. It's not about replacing our research skills, but about becoming smarter researchers.

Give it a try! Your documentation processes might become a little bit easier. 🙂