Five ways to use AI in technical writing (beyond the obvious)

As a research tool

You can retrieve code changes from a repository and use AI to look at the changes and summarise them as release notes or change note or make recommendations on what needs to be amended within the user documentation.

A screenshot of a professional software application's changelog generator interface. The UI is clean and minimalist with a white background and grey accents. On the left side, a sidebar contains date range inputs for 'START DATE' and 'END DATE' set to March 2026, a 'Generate Changelog' button with a GitHub icon, and a history of recently generated logs. The main content area displays a generated changelog titled 'Changelog' for the period of 10/03/2026 to 17/03/2026. It features a 'Download .md' button and sections for 'Release Versions' and 'New Features & Major Improvements,' detailing platform-specific updates like version numbers and new synchronization features.

Reviewing content

You can get AI to simulate different types of users and give you feedback on the content: what they think of the documentation from that persona’s perspective.

 

Results from Impersonaid user simulator app

Creating scripts to automate tasks

You can get AI to write scripts to automate mundane tasks. For example, updating diagrams or images whenever the code changes and requires a change to them.

Workflow diagram showing how AI can automatically update images

Creating management dashboards

You can create dashboards that can bring in live data to explain how the department is performing. That might be graphs for support ticket reduction for a week or a month, the most popular pages, page accuracy, and the pages with errors.

You can also create dashboards showing the return on investment for having a technical publications department.

Dashboard showing the ROI of user documentation

To help you generate deterministic tests for your documentation

This might be checking your documentation against your style guide, or testing what’s written in the documentation. If someone follows the instructions, when in the application itself, do that results in the end result that you desired.

{
"tests": [
{
"runOn": [
{
"platforms": [
"windows",
"mac",
"linux"
],
"browsers": {
"name": "chrome",
"headless": false,
"viewport": {
"width": 1180,
"height": 1480
}
}
}
],
"steps": [
{
"description": "Go to the specified URL",
"goTo": "http://localhost:8080/watson_and_holmes_intake_form.html"
},
{
"description": "Type in the First Name",
"find": {
"elementText": "First Name:",
"click": true,
"type": "Alphie"
}
},
{
"description": "Type in the Last Name",
"find": {
"elementText": "Last Name:",
"click": true,
"type": "Betaux"
}
},
{
"description": "Capture a screenshot of the completed form.",
"screenshot": "./output/form-filled-1-first-and-last-name.png"
},
{
"description": "Pause long enough to show the changes.",
"wait": 10000
}
]
}
]
}
An example of a web form. Taken from Doc Detective

If you’re interested in learning more

If you’re interested in learning more about how you can implement these tools, feel free to get in touch with us at Cherryleaf.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.