As technical communicators, we’re no strangers to automation. We’ve been using templates, macros, content management systems, and structured authoring tools for decades. But generative AI, including tools like ChatGPT, Claude, and Gemini, introduces a different kind of automation. Rather than simply helping us manage or format content, these tools can generate, revise, reorganize, and transform the content itself.
This creates some interesting opportunities for technical communicators. It also raises important questions about which parts of our work we should automate, how we should evaluate AI-generated content, and what responsibilities human writers retain.
In my article with Susan Kelley and Bremen Vance, published in the Journal of Technical Writing and Communication, we explore these questions and introduce a practical framework for integrating generative AI into technical communication workflows.
Why Do We Need a Framework for Generative AI in Technical Communication?
Much of the discussion about generative AI in technical communication has focused on whether AI will replace technical writers, whether AI-generated content can be trusted, or whether professionals should use these tools at all.
These are important questions, but they don’t necessarily help practitioners who are already being asked to incorporate AI into their work.
What these practitioners need is a way to determine which tasks are appropriate for automation, how to use AI effectively, and how to ensure that the resulting content meets professional standards.
Our article addresses this need by combining a human-in-the-loop (HITL) model of automation with a task-based approach to writing.
HITL simply means that human professionals remain involved in directing, evaluating, and controlling automated processes. Rather than asking AI to perform an entire writing project independently, technical communicators identify specific tasks that AI can help accomplish and evaluate the results before incorporating them into their work.
This distinction matters because writing involves much more than producing text. It also involves understanding audiences, gathering information, evaluating sources, organizing content, and making decisions about what information people actually need.
AI can assist with many of these activities. However, that doesn’t mean we should automatically delegate all of them to a machine.
How Does Our Framework for Writing With AI Work?
Our framework identifies four interconnected considerations for incorporating generative AI into technical communication workflows.
1. Defining Tasks
The first step is determining what we actually want AI to accomplish.
Consider a technical communicator responsible for maintaining a large documentation library. Their work might involve gathering requirements, reviewing existing content, identifying outdated information, developing new documentation, and coordinating reviews with subject matter experts.
Some of these tasks are relatively straightforward to automate. Others require substantial human expertise.
For example, AI may be useful for comparing documentation against a style guide, identifying inconsistent terminology, or summarizing existing content. However, asking AI to independently determine whether highly specialized technical instructions are accurate is much riskier.
Our framework recommends breaking larger writing projects into individual tasks and evaluating which ones are appropriate for automation.
2. Prompting Strategies
Once a task has been identified, the next step is determining how to communicate that task to an AI system.
Prompting is more than simply asking a chatbot to produce something.
Technical communicators already have considerable expertise in describing audiences, defining purposes, establishing requirements, and specifying the characteristics of effective communication.
These same skills are useful when working with generative AI.
For example, rather than simply asking AI to rewrite a document, we might provide the intended audience, the document’s purpose, relevant style guidelines, examples of acceptable content, and specific requirements for the revision.
The article also discusses several prompting strategies, including providing examples, breaking complicated tasks into smaller steps, and refining prompts based on the results.
The important point is that effective prompting requires the same kind of intentional communication that technical communicators already practice.
3. Quality Control
Generating content is only one part of the writing process. We also need to determine whether that content is accurate, complete, and appropriate for its intended audience.
This is particularly important because generative AI can produce convincing content that contains factual errors, fabricated information, or misunderstandings of specialized terminology.
Our framework recommends several strategies for validating AI-generated content, including comparing outputs against authoritative sources, applying established standards, and consulting subject matter experts when necessary.
We also emphasize that the responsibility for evaluating AI output remains with the human writer.
An AI system can help identify potential problems in a document. However, someone still needs to determine whether those problems are real and whether the proposed revisions improve the document.
4. Scaling and Automation
One of the most promising applications of generative AI in technical communication is its ability to perform repetitive writing tasks across large amounts of content.
For example, imagine a documentation team responsible for maintaining hundreds of technical documents.
AI could assist with generating metadata, identifying inconsistent terminology, converting documents into standardized formats, or developing summaries of existing content.
These are often time-consuming tasks that follow relatively predictable rules.
Our framework suggests that such tasks are particularly well suited to automation because their outputs can be evaluated against clearly defined standards.
However, scaling automation also means scaling the potential consequences of errors. Organizations therefore need to establish appropriate review procedures before automating large portions of their documentation workflows.
What Kinds of Technical Communication Tasks Can AI Automate?
One of the most practical parts of the article is a series of tables identifying writing tasks that generative AI can support.
For technical communication practitioners, we identify opportunities involving document generation, content structuring, editing, metadata management, content maintenance, and other common documentation activities.
Some examples include:
- Generating standardized documentation from existing templates and requirements
- Reviewing documents for style, terminology, and formatting consistency
- Simplifying technical explanations for different audiences
- Creating metadata, keywords, indexes, and document summaries
- Generating changelogs and assisting with documentation updates
- Developing sample API calls or other repetitive documentation elements
- Summarizing usability research findings and organizing testing documentation
These applications are not equally reliable, and each requires an appropriate level of review.
In general, the article suggests that structured tasks with clear requirements are easier to automate than tasks involving substantial interpretation or specialized knowledge.
For example, AI may be effective at applying a style guide to a document. However, it may struggle to interpret complex usability research, evaluate highly specialized technical information, or account for subtle cultural and regulatory differences in localized documentation.
Understanding these differences is essential to using AI effectively.
What Does Generative AI in Technical Communication Look Like in Practice?
To demonstrate the framework, the article walks through an example involving the development of online help documentation.
Rather than asking AI to independently produce an entire documentation project, the writer identifies a specific content development task, provides instructions to the AI system, and evaluates the initial output.
The writer can then refine the prompt to generate additional content or improve a particular section.
In our example, this process includes developing and expanding installation documentation for a university registration management system.
The process is iterative. The writer determines what needs to be accomplished, AI generates a response, and the writer decides whether that response is useful.
If the results are unsatisfactory, the writer can revise the prompt, provide additional context, or approach the task differently.
The key is that AI becomes part of an existing writing workflow rather than replacing the workflow entirely.
What About Researchers and Teachers of Technical Communication?
Although much of the article focuses on practitioners, our framework also considers how researchers and teachers can use generative AI.
For researchers, potential applications include formatting manuscripts, maintaining consistency across research documents, creating metadata, summarizing existing material, and assisting with data visualization.
For teachers, AI can help develop assignment sheets, organize instructional materials, generate assessment rubrics from established criteria, and adapt content for different learning needs.
These applications are particularly interesting because researchers and teachers often spend substantial time developing and maintaining documents that follow established conventions.
However, the same principles of human oversight apply.
Researchers must verify the accuracy of AI-assisted analyses and follow the AI policies of their publication venues. Teachers need to ensure that instructional materials support meaningful learning outcomes and accurately represent the concepts being taught.
In both contexts, the goal is to reduce repetitive work without compromising the intellectual work that makes research and teaching valuable.
How Can You Start Using Our Framework?
If you’re interested in integrating generative AI into your own writing workflow, I would recommend starting with a task that you already understand well.
Rather than attempting to automate an entire project, identify a recurring activity that consumes a considerable amount of time.
Then consider a few questions.
Can the task be defined clearly? Can you provide the AI system with examples or standards? Can you independently evaluate the results? Does automating the task actually save time after reviewing and correcting the output?
If the answers are yes, you’ve probably identified a reasonable starting point.
From there, you can experiment with different prompting strategies, evaluate the results, and determine whether the process is worth incorporating into your regular workflow.
Our framework is not intended to be a universal set of rules. AI tools are constantly evolving, and tasks that are difficult to automate today may become easier in the future.
Instead, the framework provides a series of considerations that technical communicators can use to make informed decisions as these technologies change.
What Does This Mean for the Future of Technical Communication?
One of the central arguments of our article is that generative AI does not eliminate the need for technical communication expertise.
Instead, it changes how that expertise is applied.
Technical communicators are increasingly being asked not only to create content but also to determine how AI should participate in content production, how outputs should be evaluated, and how automated processes should fit into larger organizational workflows.
These responsibilities involve many of the same skills our field has always emphasized: audience analysis, information design, content strategy, usability, research, and professional judgment.
There are also legitimate concerns about the technology itself. As we acknowledge in the article, AI is not neutral. These systems reflect the decisions, assumptions, and values of the people and organizations that develop them.
That means technical communicators should not simply accept AI tools as they are. We should also participate in shaping how they are designed, evaluated, and implemented.
Generative AI is changing the work we do.
The question is how actively we want to participate in shaping that change.
Read the Full Article
Our article provides a more detailed explanation of the human-in-the-loop framework, examples of AI-assisted writing workflows, prompting strategies, quality control recommendations, and tables identifying tasks that practitioners, researchers, and teachers can automate.
Published in the Journal of Technical Writing and Communication, 56(2), 111–144.
More Research on AI in Technical Communication
If you’re interested in how organizations can manage AI-assisted documentation workflows, you may also want to read my more recent work with Jackie Damrau and Angel Mancio.
In AI Governance Readiness in Documentation Teams: Survey-Based Insights Toward Governance Readiness, we examine responses from 90 technical communication practitioners and identify five recurring misconceptions about AI adoption, including assumptions about accuracy, automation, prompting, and efficiency.
Together, these articles explore two related questions: how technical communicators can integrate AI into their work and how organizations can develop the processes needed to use it responsibly.
Cite the Article (APA)
Getto, G., Kelley, S., & Vance, B. (2026). How to write with GenAI: A framework for using generative AI to automate writing tasks in technical communication. Journal of Technical Writing and Communication, 56(2), 111–144. https://doi.org/10.1177/00472816251332208
Learn about Mercer’s M.S. in Technical Communication Management
Mercer’s fully online M.S. in Technical Communication Management helps working professionals develop advanced skills in content strategy, usability, instructional design, project management, and emerging technologies such as AI.
If you’re interested in moving into leadership roles involving technical communication, content operations, or AI-assisted workflows, the program provides opportunities to develop the research, design, and management skills involved in this work.