What Can AI Do — and What Can't It? The One-Second Rule Every Business Owner Needs
Most business owners aren't against AI — they just don't know what it can do. Andrew Ng's one-second rule is the simplest framework I've found for answering that question.
Most business owners I talk to aren't against AI. They just don't know what it can actually do.
"I know ChatGPT is powerful, but I have no idea how it applies to my business." "I see other people using it, but I haven't had time to look into it." "I just worry that one day I'll wake up and every competitor is running on AI, and I'm still doing everything by hand."
That feeling is real — and it's not a lack of ambition. It's a lack of a clear framework.
So here's the simplest tool I've found for answering the question: can AI do this or not?
How Do You Know If AI Can Do Something?
Andrew Ng — Stanford AI professor and founder of Google Brain — shared a heuristic in his course that I think is the clearest explanation of AI's capabilities for non-technical people.
He calls it the one-second rule.
If you can do a task yourself without much thinking — in about one second — there's a good chance AI can do it too, and probably do it at scale.
A few examples to make it concrete.
Look at a photo and tell whether it contains a car. One second. AI can do that — and it's exactly what powers self-driving vehicles.
Listen to a sentence and transcribe it. One second. That's why voice-to-text is so accurate today.
Read a loan application and flag whether all the required fields are filled in. One second. AI handles this kind of document checking well.
Now flip it. Can you read a medical textbook chapter and immediately diagnose a patient? No — that takes years of training. Can you listen to someone for 20 minutes about their life situation and know exactly which visa pathway suits them? Not in a second. These tasks require experience, judgement, and nuanced understanding that AI can't replicate — at least not yet.
The one-second rule isn't perfect, but it gives you a fast, intuitive way to scan your own workflow and figure out what's worth automating.
What Can AI Actually Do for Australian Service Businesses?
Let me be concrete about what this looks like in practice.
Things AI can do right now:
A client emails you. Figuring out whether it's an enquiry, a complaint, or a follow-up takes about one second. So AI can route and categorise your inbox automatically.
Checking whether a visa application has all required documents ticked off — one second per field. AI can audit checklists without getting tired or missing items.
Answering common questions like "how long does this type of visa usually take" — one second. AI chatbots handle these well and free up your time for complex conversations.
Translating a client email from Chinese to English, or summarising the key clauses in an English contract — AI does this at a high level of accuracy now.
Things AI cannot do:
Listening to a client's full family and work history, then advising on the best immigration pathway — that's not a one-second task. It requires your expertise, your judgement, and often your experience with edge cases.
Representing a client in a complex negotiation, or adapting your strategy in real time based on how a conversation is going — AI can't replace that.
The point isn't that AI is coming for your job. It's that AI is well-suited to taking over the repetitive, rules-based parts of your day — so you can spend more time on the work that actually requires you.
Where Is AI Right Now? What the Research Actually Shows
A lot of the anxiety around AI comes from not knowing where it actually stands. Is it still early-stage? Or is it already disrupting things?
Anthropic — the company behind Claude AI — published research this year on AI's real impact on the labour market. The findings are worth knowing.
The occupations most exposed to AI include computer programmers (75% of tasks can be covered by AI), customer service representatives, and data entry workers (around 67%). At the other end, about 30% of workers — including cooks, bartenders, and lifeguards — have essentially zero AI exposure.
Here's the part that surprised most people: the workers most exposed to AI earn on average 47% more than those least exposed, and are significantly more educated. AI isn't targeting low-wage jobs — it's reshaping high-income knowledge work.
There's also a concerning signal for young people entering the workforce: those aged 22-25 are seeing roughly a 14% drop in hiring rates in high-exposure occupations. Entry-level knowledge work is getting harder to break into.
But here's the context that matters: even though 94% of tasks in computer and math fields are theoretically AI-capable, only 33% are actually being done with AI today. The technology is ahead of adoption. That means there's still time — but the window isn't staying open forever.
The Question Worth Asking
The fear of being left behind is a valid instinct. But fear without a framework doesn't help you make any decisions.
The more useful question is: in everything you do today, what are the one-second tasks? Make that list. That's your automation backlog.
You don't need to become a technical expert. You don't need to learn to code. You just need to apply this one filter to your own workday and see what comes up.
That's exactly what we help service businesses do at AU Plus — map out the one-second tasks, then actually build the automations that take them off your plate. If you're curious about what that looks like for your business, you're welcome to book a free discovery call.