Microsoft
MVP
1997-2018
John McGhie
Documentation that serves your purpose is a piece of
engineering, not a work of art.

Availability: Retired
By Post:
Unit 38, 24-28 First Ave,
Blacktown,
Sydney, Australia 2148
By Email: mailto:john@mcghie.com.au
By Phone: +61 (0)4 1209 1410
Welcome to my personal web page. These
days I am retired, living in Sydney, Australia.
I take this opportunity to wish you
happiness and great success in your undertakings now and in the future. If you
are interested in my career the way it was, Click here
to download my former resume.
I was the principal of McGhie Information
Engineering Pty Ltd, a privately-held Australian
company specialising in documentation engineering.
Documentation engineering begins with the
premise that "Documentation
that serves your needs is a piece of engineering, not a work of art"
– which just happened to be my company slogan. That's what we really did:
We analysed what the customer wanted people to be able to do after
reading your documentation and we analysed what they already know. Then we
designed a suite of information resources to bridge the gap.
Documentation engineering is normally
practiced by a team of five or six people, including the documentation
engineer, one or more subject matter experts, a business analyst, a project
manager, an editor and a production editor.
Note that in my day, no Artificial
Intelligence of any type existed. Today, its role would be confined primarily
to research compilation.
The current AI platforms are Large
Language Models (LLMs). LLMs are very
good at predicting what “an average person” might say on any given subject —
recycling already-existing knowledge into useable text.
However, my customers required 100%
accurate, completely reliable text that does not yet exist, to describe the use
of products that do not yet exist.
Many such assignments have a safety- or
security- or finance-related exposure where mistakes can kill people or have a
catastrophic impact upon the survival of the customer’s business.
The likelihood that an AI application may
hallucinate — make up an answer it doesn’t “know” — would present an
unacceptable and unmanageable risk exposure.