In April 2024, I was the hip Senior Manager writing a blog post about AI and ChatGPT. My experience was a bit of a novelty – “write us an article Em!”. Rereading the article recently, I realised it was well overdue for an update.
I am no longer at the frontier of a new technology – AI is built into the tools we’re all using every day. It's in our inboxes, our meeting platforms and our clients' intranets. And much of our work as change and communications professionals is no longer just using AI but helping organisations and their people adjust to it.
So here's what I'd keep, what I'd change, and what I got wrong.
Then: Never put anything into an AI generator you wouldn't want on the front cover of the Times.
Now: It's still a good gut-check, but as a whole policy it's too blunt.
Many organisations now have enterprise versions of AI tools that keep information siloed or private. "Never" made sense when the only option was a free public tool. It makes less sense when you have an approved, secure one.
The better question is no longer "should I paste this in?" but "what is this information, and which tool is approved for it?"
A traffic-light approach can help:
For clients working with government or defence, add the layer of information classification and data sovereignty. Also refer to v2.0 of the responsible AI use policy published by the Digital Transformation Agency.
At Parbery, we still don't put client information into an AI program without approval. What's changed is that we ask which program, and what information, rather than relying on a blanket ban.
Then: Have a conversation. Don't expect one prompt to deliver what you want.
Now: Still true, but the skill has shifted from prompting to context.
The best results no longer come from clever one-liners. They come from giving the tool something to work with – who the audience is, what they already know, what tone suits the moment, and examples of what good looks like. Many tools now let you save that as standing instructions or a reusable project, so you don't start from scratch every time.
Think of it as the difference between briefing a stranger and briefing a colleague who has worked with you for a year.
Then: Authenticity is something AI can't produce on its own.
Now: I'd still say this, but there's a new risk – everyone is sounding the same.
When a whole industry drafts with the same handful of tools, the results converge. The same rhythm, the same "I'm excited to share", the same tidy three-part lists. Employees notice. If your change message reads like every other message they've received this year, it won't feel like it came from a leader who understands what they're going through.
Loading brand guidelines helps, but it only goes so far. The real safeguard is a communicator who knows the audience, the history and the emotion behind the words, and who is willing to rewrite the polished paragraph that says nothing.
Then: I compared AI to a calculator for the writing-challenged, it knows the most probable answer but that doesn’t mean it’s the right one.
Now: I'd retire that analogy. The tech we have now is far beyond a calculator for words, and even the best writers are using it in some form.
The problem today is not clumsy, nonsensical sentences. It's the opposite. Modern tools write fluent, confident text that can still be wrong. They agree with whatever you suggest, and gloss over the detail that matters most.
That makes reviewing harder, not easier. If you're checking anyone's work, including your own, here's what to look for:
That first one matters most for high-stakes communications such as safety, regulatory and workforce messaging. AI can help draft them. A person has to own them.
Then: Excel formulas, meeting agendas and Instagram handles.
Now: Those are all still fair game, but in change and communications work, the more valuable uses have grown:
On that last point: recording and transcribing candid workshops or focus groups is a trust issue, not just a convenience. If people are speaking openly about how change affects them, they deserve to know who, or what, is listening.
Then: I said please and thank you to my digital assistants, partly out of gratitude and partly to hedge my bets against the robot overlords.
Now: I've changed my mind, at least a little.
In 2025, OpenAI's CEO Sam Altman said on X that pleasantries add up to real computing costs at scale, and there's a wider conversation about the energy and water AI uses. The per-message impact is small and hard to measure precisely, so I wouldn't lose sleep over one polite sentence. But skipping the pleasantries when a message triggers a whole extra response is an easy change.
When I wrote the original article, AI was something communicators used. Now it's often what we're communicating about.
Many of the change programs we support involve rolling out AI tools to people who are excited, sceptical, or quietly worried about their jobs.
We’ve found communicating well in that environment looks like:
Instead of a list of tools (which dated almost immediately last time I did this) here’s a few things to keep in mind:
Common sense still prevails. It just needs updating more often than we thought.
Move from ideas to outcomes with confidence
Reach out to discuss how we can partner in designing, delivering and embedding practical solutions that create real impact from day one.
Parbery acknowledges the traditional owners and custodians of country throughout Australia, and their continuing connection to land and community. We pay our respects to their people, their culture, and to elders past, present and emerging. Parbery also specifically acknowledges and respects the contribution made to the betterment of country in the regions we operate in: ACT, Qld, SA and Vic.
© 2026 Parbery Consulting. All Rights Reserved.
Site designed and built by Shannon Towell Design
