Artificial Intelligence (AI) for Technical Writers

Lead Writer: Caity Cronkhite | Peer Reviewer/s: Kieran Morgan | Expert Review: Derek Moeller | Managing Editor: Kieran Morgan

This chapter examines the significant impact of artificial intelligence (AI) and large language models like ChatGPT on technical writing. It highlights AI’s role in enhancing productivity by aiding in routine writing tasks and discusses its limitations in handling novel concepts. The chapter covers AI terminologies, its integration into writing processes, the importance of security and privacy in professional settings, and the role of technical writers in producing original, user-centric content.

Audience Icon Who Should Read This
• Aspiring Technical Writers
• Beginner Technical Writers
• Career Advancers
• Managers of Technical Writers
• Cross-Domain Professionals
• Consultants
Table of Contents: Technical Writing Process
Previous: Chapter 6: Career Flexibility in Technical WritingNext: Chapter 9: Collect Information

1. Introduction

In late 2022, OpenAI released the first publicly available version of ChatGPT, an interface to a large language model (LLM) designed to generate human-like text. Very quickly, this groundbreaking technology took the world by storm. In a matter of weeks, more than one million people downloaded ChatGPT, making it the “fastest-growing user application in history,” according to research.1

Recent advances in artificial intelligence technology, particularly in large language models, have enormous implications for writers and communicators of all kinds. Humans are already leveraging artificial intelligence to write outlines, academic papers, blog posts, journalistic articles, and even full-length books. What impact, then, can we expect it to have on our roles and careers as technical writers?

In this chapter, we demystify some fundamental concepts of artificial intelligence and terminology. Through a worked example, we demonstrate how useful ChatGPT can be in streamlining typical writing tasks. For those wishing to explore more advanced uses, such as understanding the tool’s limitations, finding solutions to typical issues, and applying ChatGPT to numerous other technical writing scenarios, we recommend our comprehensive resource: ChatGPT Prompt Library for Technical Writers.

Artificial intelligence isn’t just a disruptive fad. When leveraged skillfully, it can be a powerful tool to boost productivity, eliminate the drudgery of some repetitive tasks, and increase your own efficiency as a technical writer.

2. [Theory] What Is a Large Language Model (LLM) and How Does It Work?

Artificial intelligence refers to the simulation of human intelligence by machines and computer systems. These computer systems are trained on massive amounts of data to simulate the process of learning, reasoning, problem-solving, perception, language, and decision-making.

Generative artificial intelligence is a subset of artificial intelligence models and the most recent major advancement in artificial intelligence technology. Generative artificial intelligence models are capable of generating new content—such as images, text, or music—that isn’t directly copied from training data.

A large language model, or LLM, is a subset of artificial intelligence technology that’s specifically designed to parse, understand, and generate humanlike text and content. Large language models are built on a deep-learning architecture and are trained on massive amounts of text data, including books, articles, websites, and, yes, even technical documentation.

You can interact with a large language model by entering a prompt into a platform or app, such as ChatGPT, that uses this technology. A large language model is a generative artificial intelligence model: it analyzes the context and patterns it learns from training data to predict the most appropriate next words or sentences, then generates a response that sounds like natural language.

What Does That Mean Icon What Does That Mean?
Artificial Intelligence (AI)
The simulation of human intelligence by machines and computer systems.

Generative Artificial Intelligence
A subset of artificial intelligence models capable of generating new content—such as images, text, or music—that isn’t directly copied from training data.

Large Language Model (LLM)
A subset of generative artificial intelligence models designed to understand and generate humanlike text.

A large language model developed and released by OpenAI, designed to generate humanlike text and content in response to human prompts.

2.1. Dos and Don’ts for Using Artificial Intelligence at Work

Artificial intelligence tools have tremendous potential to help technical writers streamline their workflows and processes. However, tread carefully: the technology is still new, and many companies limit or prohibit the use of large language models for work purposes. Large language models, such as ChatGPT, store your input data for training purposes by default—although there are options not to do so, by purchasing premium or enterprise-grade plans. Here are a few dos and don’ts to keep in mind before you start using large language models in your own career.

Get permission from your employer before you use any large language model to write proprietary content.Use public-facing large language models to write proprietary content unless you have express permission from your employer.
Review and fact-check content for accuracy and truthfulness—and to avoid plagiarism and bias—before publishing.Include sensitive information, like trade secrets or customer data, in your prompts.
Make sure that using artificial intelligence in your work complies with legal regulations and data protection laws in your industry.Publish content that was written by a large language model without human review for accuracy, truthfulness, and data compliance.

3. [Opinion] Is Artificial Intelligence Going to Take My Job?

Artificial intelligence technology, particularly large language models, is already dramatically changing the writing process and will continue to do so as the technology evolves. As we discuss in the next section, Chapter 8: Artificial Intelligence (AI) for Technical Writers > 4. [Example] Using ChatGPT to Document a Payment Software API, there are several ways that technical writers can leverage large language models to streamline common aspects of the writing process, from writing outlines to applying style-guide standards to our content. But can artificial intelligence replace technical writers completely?

Not so fast. Remember what we learned about large language models? Large language models are generative artificial intelligence models, which means they learn from existing content to predict what might come next in a sentence or paragraph.

A large language model can easily produce information about products and concepts that have been written about extensively, assuming that information has been incorporated into the data used to train the model. However, a large language model can’t formulate content or responses around ideas that have never been written about or incorporated into its training data set—at least not yet!

As technical writers, we’re typically tasked with writing about new products, concepts, ideas, and technologies—a task that current large language models aren’t programmed to do. In fact, the content that technical writers produce is likely to become even more important as artificial intelligence and large language model technology becomes more ubiquitous at work. The content we write about new products, services, and technology will likely become part of the corpus of information on which large language models are trained about new ideas. Large language models can’t make decisions about what kind of content users might need, nor can they manage or maintain our content library for us.

Additionally, many technical writers create content for highly sensitive or proprietary products, data, or industries. Artificial intelligence and large language model technologies are relatively new in the public sphere, and there are widespread concerns about the security and safety of using these products to write content about highly sensitive information. Because most publicly available artificial intelligence products don’t have established standards for protecting the data they receive in prompts,2 many companies and industries shouldn’t use artificial intelligence or large language models to generate content at all at this time.

Artificial intelligence will certainly change how we do our jobs as technical writers, but it isn’t going to replace us, at least not for now. By embracing this new technology and incorporating it into our work, we can streamline the technical writing process to make our work less tedious and produce better content for our users. Let’s explore how.

4. [Practice] [Example] Using ChatGPT to Document a Payment Software API

Let’s take a look at a scenario where artificial intelligence might help us streamline various aspects of the technical writing process.

Imagine you’re a technical writer working for an enterprise technology company that builds payment software. Your job is to document the company’s APIs for external developers. You’re the only writer assigned to the job, so you decide to use ChatGPT to help you streamline your writing process and save time.

Note Icon Note
A Note on Artificial Intelligence Preferences
ChatGPT isn’t the only artificial intelligence large language model available on the market today. However, because of its ease of use and general accessibility, we refer primarily to ChatGPT in this chapter and elsewhere in this book.

4.1. Getting Started: Setting Up ChatGPT for Success

First you need to tweak a few settings to make sure that ChatGPT produces content that matches your audience’s needs and that applies technical-writing industry standards and best practices. To increase our chances for success, we use ChatGPT’s custom instructions feature.

Custom instructions allow us to:

  • Provide information about who we are, our role, and the kind of content we’re working on.
  • Define the audiences we’re writing for.
  • Include specific guidelines and instructions for the content ChatGPT produces, such as voice and tone, punctuation, or style guidelines.

Here’s an example of the custom instructions you might use for your technical writing role:

4.2. Use Case 1: Content Planning

Now that you’ve set your custom instructions, it’s time to get to work planning the documentation for a new API your company is working on. You want to outline the structure and information architecture of the documentation before you start writing.

With a simple prompt (and the custom instructions you provided earlier), ChatGPT produces an outline of basic topics you’d likely want to include in API documentation for a feature like this.

You also want to provide some useful examples and developer tutorials. ChatGPT can help with that too!

4.3. Use Case 2: Writing Content

Planning content is great, but you have deadlines to meet, and you need to start writing.

Remember, ChatGPT only has access to information and data that’s publicly available on the web. Your company’s API hasn’t been released yet, so ChatGPT can’t write factual information about your API and its features. However, if prompted, it will certainly generate plausible-sounding—and possibly wildly inaccurate—content!

Luckily there are a few topics in your documentation outline that are more generic, such as Section 6: Securely Handling Credit Card Data. You use ChatGPT to write a first draft of a topic that describes what PCI DSS is and why it matters. Here’s a snippet:

Not a bad start!

(Keep in mind that, although that first draft looks pretty good, you always need to verify the information to make sure it’s accurate and factual before you publish to your audience.)

Tip Icon Tip
Leverage ChatGPT’s Technical Writing Abilities
Many folks aren’t aware that GPT-4 can translate code into human language explanations—with some limitations. This capability can be extremely useful for technical writers working on developer documentation. For instance, an API’s code endpoints, complete with type hints and function docstrings, can be fed into GPT-4, and it will produce a rough draft explanation of the API, complete with prompts. Because the accuracy of ChatGPT’s explanations varies, the rough draft must be reviewed and validated by subject matter experts—a process that you, as the writer, are responsible for orchestrating. Before doing so, remember you should never put your organization’s code into ChatGPT without first getting permission from your employer.

4.4. Use Case 3: Content Editing, Formatting, and Conversion

Your company publishes documentation in Markdown, so you need to convert the content ChatGPT wrote for you into the Markdown format. Never fear: ChatGPT is adept at that too.

In less than a minute, ChatGPT tags the content accurately, allowing you to copy and paste it seamlessly into your existing Markdown content.

5. [Template] ChatGPT Prompt Library for Technical Writers

ChatGPT Prompt Library for Technical Writers

  1. Wodecki, B. (2023, March 1). UBS: ChatGPT may be the fastest growing app of all time. ↩︎
  2. Divatia, A. (2023, June 22). How companies can use generative AI and maintain data privacy. Forbes. ↩︎
Would love your thoughts, please comment.x