Skip to main content

AI policy

As a technical writer, I often write documentation that acts as the source material for both human users and AI tools.

I am a user-focused technical writer, and I take a human-first approach to technical writing. My professional interests lean toward human user experience, through the principles of information architecture and cognitive accessibility.

An LLM can understand anything written for a human, but a human can't necessarily understand everything written for an LLM.

Using AI​

I use AI tools sparingly, and only where I believe it can genuinely provide value and quality. I have the final input on everything I produce, to ensure that your content is clear, accurate, and high quality, and that it meets the highest ethical and legal standards.

If you have an existing AI policy, I'll work with that. If you don't yet have an AI policy, I'm happy to help you develop one while we work, so we can determine how and when to use LLMs for drafting and editing content, and whether to feed commercially sensitive information about your product or company into AI tools.

For example, I might use AI tools for:

  • Sourcing general information about the subject matter (e.g., "Find me an article on how a clutch plate works").
  • Verifying general subject matter knowledge (e.g. "This is my understanding of how a clutch plate works; are there any inaccuracies?")
  • Checking document style and consistency (e.g. "Make sure I've used clutch plate instead of clutch disc throughout.")

I request and check sources for all information, and I fact-check all technical information with human subject matter experts or human-written resources.

I don't use AI for:

  • Learning or writing about your specific product (e.g. "My client is making a special type of clutch plate that lights up; why is this useful?" I want to work from your description of your product and the problem it solves.)
  • Making decisions about structure and scope (e.g. "Which information does the user need first, and why?" This is where human decision-making is essential!)
  • Writing final documentation (e.g. "Here are some notes, please generate a full page of documentation." Docs are more than a page at a time, and LLM-written docs don't take into consideration the big-picture information architecture of the docs.)

I'm not interested in any role that primarily involves babysitting an LLM.

Ethical disclaimer​

I have ethical concerns about LLMs and generative AI tools, including their technical inaccuracy, their data training ethics, and their environmental impact. I also have concerns about their impact on human cognitive functioning, creativity, and mental health.

My own experiments indicate that AI tools produce technical content that looks like documentation, but which doesn't act like it. It often doesn't provide the level of detail, certainty, or crunchy specificity that users are really looking for, and it requires more editing time than its human-written equivalent.

However, some AI tools do have valuable applications, and it's not realistic to completely reject them in the current tech landscape. It's a difficult balance. Like everyone else, I'm still figuring it out.

I'm open to discussing specific approaches with clients who have their own requirements around AI use. I hope my current policy provides reassurance that I use AI tools thoughtfully and intentionally.