ChatGPT Upgraded Your Free Account and Did Not Tell You
ChatGPT upgraded your free account this month and did not tell you
Hello, and welcome to this week's edition.
The biggest story for small business owners this month is one nobody announced to you directly. In mid August, OpenAI moved free and Go users of ChatGPT onto a newer default model. Google shipped a faster Gemini model on 13 August. If you tried an AI tool last year, decided it was underwhelming, and never went back, the thing sitting in your browser today is not the thing you tested.
Last week we looked at what changed in AI for small business. This week the theme is narrower and, I think, more useful: the tools quietly got better, and almost nobody changed how they work because of it.
That gap is where the money sits.
The tool improved. Your workflow did not.
Here is the pattern I see over and over. An owner opens ChatGPT, types a question, gets a decent answer, and closes the tab. Nothing about how the business runs on Tuesday is different from how it ran on Monday.
A sharper model pointed at an undocumented process produces an undocumented process. Faster.
The numbers back this up. Australian small and medium businesses are not avoiding AI. They are using it without a structure around it.
69% of Australian SMEs reported regular AI use in January 2026, up from 40% in July 2024.
That is a fast climb. But adoption is not the same thing as return. Using a tool is not a system.
And the businesses standing back are not standing back because AI is too hard.
Around 65% of Australian businesses not adopting AI cite distrust in AI decision making, or a preference to keep humans in control of their processes.
Read that again. The barrier is not the technology. It is control. Owners do not want a black box making calls inside their business, and that instinct is correct.
The fix is not more AI. It is a documented process with a named human at every decision point. Once that exists, AI becomes a tool you can supervise instead of a risk you are carrying.
Five tools worth ten minutes this week
None of these need a technical background. None of them require you to change how your business runs on day one.
ChatGPT, free tier
Now running a newer default model for free and Go users. Best used for first drafts, summarising a long email thread before you reply, and thinking out loud before you write something that matters.
Google Gemini
Sits inside Gmail and Docs. If your business already runs on Google Workspace, there is nothing new to log into and nothing new to learn.
Canva Magic Studio
Resizes one graphic across every social format and strips backgrounds without a designer. It removes the specific friction that stops most owners posting at all.
Notion AI
Turns a messy meeting note into a list of actions with owners against them. Useful for the jobs that currently live only in your head.
Motion
Rebuilds your calendar around what actually has to happen today. This one is paid, and it is worth it only if your diary is genuinely the bottleneck. If your bottleneck is quoting, buy something else.
The real question is never which tool. It is which task. Pick the task first, then go looking.
The quick win: time it before you automate it
Thirty minutes this week. No new software, no subscriptions.
One. Pick the admin task you repeat most. Quotes, invoice chasing, booking confirmations, whichever one you dread on a Monday.
Two. Do it once with a timer running. Write down every step, including the ones you do without thinking.
Three. Multiply those minutes by how often you do it in a month. That number is your business case, and it is usually bigger than owners expect.
Four. Cross out any step that exists only because nobody ever wrote the process down. Most lists lose two or three steps right here, before any tool is involved.
Five. Now open an AI tool and hand it the shortened list. You are automating a process instead of a habit.
This is the whole method in miniature. If you want the longer version, I have written it up in the full guide on how to implement AI in your small business.
Workflow before automation. Always.
Watch out for: three things costing owners real money
This part is less fun, and it matters more than the tools.
Deepfake fraud reportedly cost businesses around 96 million US dollars in the first four months of 2026, following roughly 1.1 billion US dollars in losses globally across 2025.
Small businesses are hit harder than large ones. Not because owners are careless, but because approvals in a small business happen on trust and a phone call rather than a documented check. That is a systems gap, and attackers price it accordingly.
One. The voice that sounds like you
A few seconds of your voice from a webinar, a podcast or a video post is enough to clone it. Your bookkeeper then takes a call from you approving a payment or a change of bank details. Bad grammar and robotic delivery no longer give it away.
Do this: make it a standing rule that no payment and no bank detail change is ever approved on a call alone. Ring back on a number you already had, not the one that just rang you.
Two. The supplier who never existed
A clean, plausible invoice from a new vendor clears a light onboarding process and gets paid before anyone checks the entity is real.
Do this: verify new supplier bank details against a phone number from their own website, never the number printed on the invoice.
Three. Your client data inside a free chatbot
The Australian Signals Directorate has flagged data leaks, privacy breaches and unreliable outputs as live risks for small businesses using cloud based AI tools.
84% of Australian small businesses reported being hit by a cyber incident in 2026, up three percentage points on 2023.
Do this: check whether the tool trains on your inputs before you paste anything into it. Contracts, bank details and client personal information do not belong in a free consumer chatbot.
The big picture: the question changed while you were busy
For three years the industry argued about whether small businesses should use AI at all. That argument is finished. Nearly seven in ten Australian SMEs already use it regularly.
The live question in 2026 is narrower and considerably harder. Where exactly does a human stay in control?
You can see what the answer looks like at scale. UPS put AI agents onto its customs documentation process this year and lifted the share of small packages clearing customs same day without manual handling.
UPS moved same day customs clearance without manual intervention from 21% to 97% by putting AI agents on a documented process.
Notice what came first. The process was already mapped. The AI was applied to something that had a defined shape, defined inputs and a defined exception path. Nobody pointed a model at a mess and hoped.
AI did not get harder to use this year. It got harder to supervise.
That is not a tool problem. It is a systems problem. And it gets solved on paper before it gets solved in software.
What I would do this week
If you take one thing from this edition, take the timer exercise. Thirty minutes, one task, one complete list of steps. You will find the automation candidate yourself, and you will find two steps that should not exist at all.
Then set the payment verification rule with whoever handles your money. It costs nothing and it closes the gap the deepfake operators are actually walking through.
If you would rather have someone map it with you, that is the work I do. I am based in Wollongong and work with businesses across the Illawarra and around Australia. You can see how that engagement runs on the AI automation Wollongong page.
See you next Monday.
Structure, not AI noise.
Ashley Ruperto, Limitless Digital Co
