After my experience with Copilot while writing a blog, I wanted to find out how useful the tool would be for more serious work. So, I used it to convert an operating manual into a system description. The starting point seemed promising: a complete manual as input and a clear template as the basis.
From lots of input to very little output
In practice, the result was disappointing. The structure of the template was followed neatly, but the content itself remained thin. Instead of coherent paragraphs, I got a collection of disconnected sentences. It was only when I broke the assignment down into individual chapters or sub-sections that the output became genuinely useful.
This experience highlights something important: Copilot struggles with large, complex tasks. It works less like an all-knowing solution and more like a colleague. Give it a clear, well-defined task, and it can be a valuable assistant. Make the assignment too large or too vague, and the result tends to remain superficial.
The more specific your request, the better Copilot performs
The quality of the output depends heavily on your prompt. It helps to clearly specify what Copilot should do, in what role, in what format, and for which audience. For example, you can provide a task, context, and clear requirements for the output.
A prompt might look something like this:
As a technical writer, I need to create a user manual for a coffee machine. I already have the installation manual for this machine. Based on that document, write the introduction for the user manual and follow the structure provided in the template. The output should be suitable for inclusion in a Word document. Please first confirm whether the installation manual provides sufficient input.
AI makes your work faster, but not automatically better
For organizations, having an extra "colleague" that helps complete tasks more quickly and efficiently is an attractive prospect. However, those benefits do not come automatically. Users still need to provide effective prompts and remain responsible for the final content. AI-generated output can be incomplete or slightly inaccurate, which means it should always be reviewed critically.
The real value of AI therefore lies not only in generating content, but also in the combination of generating, evaluating, and refining it.
What have you learned from using AI in your work? I'd love to hear your experiences.
Kind regards, Birgit
birgit.maas@eluxis.com
LinkedIn: Birgit Maas