How AI Will Change Technical Content: Part 1

This post is the first post in a 4 part series on how AI will change technical writing in the coming years. It's intended to help individuals and teams think about how to use AI in their day-to-day workflows.

The origin of this series is from many conversations with folks doing work in the technical content space. They're in developer relations, product marketing, and in technical writing or documentation engineering.

Before we dive into the details of the first stage, let's briefly outline the four stages of AI maturity in technical content:

  1. Individual Experimentation
  2. Assisted Automation
  3. Orchestrated Automation
  4. Full Automation

The focus of this post is on individual experimentation.

Stage 1: Individual Experimentation

In this initial stage, technical writers and developers are just beginning to explore the potential of AI in their work.

It's experimental.

These are the folks that are willing to deal with the inadequacies of solutions and just want to "see it for themselves".

They're experimenting with various tools and techniques to enhance their personal productivity and improve the quality of their documentation.

Examples of applications

Let's look at some examples of how individuals might be using AI in this stage:

1. Grammar and Style Checking

Many writers are turning to AI-powered grammar and style checkers to improve their writing. Tools like Grammarly or Hyperlint use natural language processing to identify not just spelling and grammar issues, but also suggest improvements in style, tone, and clarity.

Now this is all manual, there's no automation behind it. As one person put it:

"I just copy and paste it into ChatGPT, just to see what it says."

2. Content Summarization

AI-powered summarization tools can help writers quickly distill key points from lengthy documents or research papers. This can be particularly useful when updating docs to reflect new research or technical advancements.

A developer might use a tool like Claude to generate a quick summary of a lengthy technical specification before incorporating the information into their documentation.

As one interviewee put it:

"I didn't have time to listen to the whole podcast, so I threw it into Gemini and got a summary of the key points."

3. Code Documentation Assistance

Developers are exploring AI-powered tools to assist with code documentation - it should come as no surprise.

Tools like Cursor and GitHub Copilot generate initial drafts of code comments or function descriptions based on the code itself. While these still require human review and editing, they can significantly speed up the documentation process.

Developers described this as a double-edged sword because AI frequently lacked the context on why a change was made - not just what that change is. Others mentioned that maintaining docstrings was also challenging.

4. Translation and Localization

For teams working on multilingual documentation, individuals are experimenting with AI-powered translation tools. While these aren't perfect, they can provide a starting point for localization efforts, potentially saving time in the initial translation process.

The big danger with localization is:

  1. maintenance and
  2. quality.

Most teams we spoke to worked with 3rd party providers to help ensure that translations were good. This is a fast moving space and we found that projects where this was an experiment were often given up.

5. Content Suggestions & Generation This is a bit of the kitchen sink, but almost all we spoke to had at least experimented with using AI to generate new documentation and content.

Most we spoke to found that the output quality just was not good enough for them to use outside of manual processes.

Challenges and Considerations

While individual experimentation with AI tools can lead to exciting discoveries and productivity gains, it's important to approach this stage with a critical mindset:

  • Quality Control: AI-generated content or suggestions should always be carefully reviewed for accuracy and appropriateness. In the AI community, this is referred to as "slop".
  • Consistency: Individual use of AI tools can lead to inconsistencies across a documentation set if not carefully managed.

This comes up repeatedly with documentarians.

When you've got a lot of individuals experimenting what happens is that each of them experiment in their own way. The big problem with that is that your documentation ends up in a very inconsistent state. There is different uses of language there's no consistency and style. It's really troublesome.

  • Privacy and Security: When using third-party AI tools, consider the sensitivity of the content being processed and ensure appropriate safeguards are in place.

While this was a minor point, some companies expressed a preference for hosting and managing their own AI solutions to maintain control.

Looking Ahead

As individuals become more comfortable with AI tools and discover effective ways to incorporate them into their workflows, organizations may begin to standardize on certain tools or practices.

In the next blog post on this series, we're going to be diving into maturity level 2 - assisted automation.

Conclusion

The individual experimentation stage is an exciting time for technical writers and developers.

It offers the opportunity to explore new tools, enhance personal productivity, and potentially discover innovative ways to improve documentation quality.

At Hyperlint, we believe strongly that AI will benefit the technical writing and communication communities.

By starting with individual experimentation, we lay the groundwork for more advanced stages of AI integration in the future, always keeping in mind that the goal is to enhance, not replace, human expertise in technical communication.