Software development is changing around us, and it’s changing your career

GitHub Copilot can today explore a repository, plan a change, edit code on a branch and open a pull request. This is one sign of how software development is changing. A coding agent can now take on several steps between someone describing a problem and a developer reviewing a proposed solution.

AI-assisted development can also change the build-or-buy decision for organisations. A team that previously subscribed to a SaaS platform might now use a coding agent to build a small application around its own processes. It still has to take responsibility for maintenance, performance and security. It also inherits work a supplier might otherwise do, such as explaining how to use the application, training new starters and giving support staff enough information to troubleshoot problems.

That also includes documentation.

Let’s imagine a team builds its own application for approving purchase requests. The first version works because everyone involved knows who can approve what. Six months later, a new manager needs to delegate approvals while on leave, and nobody can explain what happens to requests already in the queue.

That gap might call for a better documentation. It might also reveal a decision the team never made about how delegation should work.

That’s an opportunity for technical writers to get involved earlier in the development process.

The questions we ask while documenting a feature can help a team decide how it ought to work while there’s still time to change it.

Bringing documentation into the pull request

A coding agent proposes a change to purchase approvals. A technical writer uses AI to identify affected guides and tests the change. The product team decides what happens to queued requests before the software and documentation ship together.

A pull request gives technical writers a concrete point at which to get involved.

For example, if a proposed change alters who can approve a purchase request, a writer could use an AI assistant to compare the change with the user guide, onboarding material and support articles. you could ask it to point to the evidence for each suggested update and flag any questions the available sources leave unanswered.

The writer can then check the behaviour in a test environment and bring unresolved questions back to the team. For example:

  • What happens to requests submitted before the change?
  • Will users know who can approve them?

The answers could affect the interface, the release notes, and the guidance people follow on day one.

To be able to do this, technical writers need to master AI. In practice, that means knowing what to delegate, which sources to trust, how to verify the output and when a question needs a decision from the product team.

What would you do with the time AI saves?

On I’d Rather Be Writing blog, Tom Johnson posted on how he compressed five days of routine documentation work into one. There were trade-offs. He maintained release coverage while pausing deeper work such as restructuring the docs. His proposal is to spend some of the time freed up by automating recurring work on making documentation easier for AI agents to use, then bring evidence of where those agents struggle back to the product team. See: A three-phase path to reinventing the tech comm career.

Professor Ethan Mollick made a broader argument in One Useful Thing in 2023. When technology is changing fast enough to make career predictions unreliable, it pays to experiment and stay flexible. See: Becoming strange in the Long Singularity.

For a technical writer, that could mean starting with one recurring task, such as checking which pages a pull request affects. Compare the AI-assisted result with your usual review. How much time did you save after checking the output? What did each review miss? If the workflow frees up time, decide what to spend it on.

Build your AI skills with Cherryleaf

If you want to use AI in your day-to-day technical writing, our ISTC-accredited course covers practical ways to use it across the documentation process. Demonstrations include using repository changes to identify updates to a user guide and working with OpenAPI files to create developer content.

See:  Using Generative AI in Technical Writing.

If you’re responsible for how a team adopts AI, our programme for documentation managers and senior writers takes you through designing and testing a workflow. You’ll define what success looks like, decide where human approval belongs and use the results to recommend whether the workflow should go ahead.

See: Managing Documentation Projects with AI.

Leave a Reply

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