Time Saved by Using AI to Perform a Common
Assignment in Technical Writing Courses

Michael Alley, Andrew Abrantes, and Leo Randolph
24 April 2026*
 Picture
   
Introduction

This page presents an experiment to determine the minimum time required to produce an excellent version of a common assignment in technical writing courses. By the word excellent, we mean a level that would earn an A in the course and use validated sources.
For the past 20 years, a common assignment in one of the largest technical courses in the United States [1] is to describe a technical process to a nontechnical audience [2]. Given that so many students are using artificial intelligence (AI) to perform assignments, the question arises whether using AI could assist in helping create an excellent version of this process description and the minimum time that the author would need to invest for that level of quality.


Methods

This section presents the process that we adopted for using AI to create a process description. Our overall process for doing the experiment was for each of us to create a process description. Using the assignment of the technical writing course (English 202C), we selected a target length of 1000 words, plus or minus 100 words. 

Assumptions. The following are assumptions that we made at the outset.
  1. quality of product has to match that of assignments that earned A's in the course
  2. author is confident that description does not contain any hallucination
  3. author is confident that proper credit is given to appropriate sources
These assumptions were used to determine whether an author would perform or assist in a step in the writing process. If AI was 





Selecting the Topic. Text
1. chose topic that I knew something about
2. a topic to be interesting to me but also the reader (nontechnical writing instructor)
3. a topic that would fulfill the assignment
4. a topic that will be new for the instructor
Assumption: thinking about topic 
One thing is that we could ask AI what it thinks


Researching and Drafting.

Incorporating Graphics.

Revising and Drafting. 
Results

Presented here will be the three descriptions

​Andrew tested Gemini, Copilot, and Claude. Copilot was the weakest, but could not compile visuals. Microsoft Copilot could not provide a Microsoft Word document.
    Gemini gave something quickly--about 1000 words. Provided a memo header. It could not give images. It was comparable 
    Claude was the slowest to produce a document, but the document was the strongest output and included a 

Leo using ChatGPT free

Michael using ChatGPT Plus and Claude
Picture
​References
  1. Michael Alley, The Craft of Scientific Writing, 4th ed. (New York: Springer, 2018).​​
  2. Michael Alley, "Section 1: Grammar," Writing as an Engineer or Scientist (1997).​
​
* Except for the AI example in Figure 1, AI was not used to draft or revise this webpage.