What AI for an SME actually means.
AI for an SME in Hong Kong or Australia means replacing work that still lives in WhatsApp, Excel, or paper with systems and agents that run on your own infrastructure. Do a readiness audit before you buy an agent. Then build the stack, write the rules, and train the people who own them.
The WhatsApp, Excel, Paper test
Three questions before you spend a dollar on an AI agent. Yes to any of them means infrastructure comes first.
Does critical work live in chat threads? The order that lives in a conversation is invisible to any system and to any agent you buy.
Does the business depend on a spreadsheet only one person understands? Macros and #REF! errors are not a database, and agents cannot act on them reliably.
Does anything important exist only in a filing cabinet? An agent cannot read what was never digitised, and neither can your next hire.
Do this before you buy an agent
Walk one order, one invoice, and one customer enquiry through the business. Note every time it touches WhatsApp, a spreadsheet, or paper.
An honest pass across the business: what is genuinely worth automating, what is not, and what to ignore. One audit at an Australian manufacturer produced three builds.
Can your data leave the building? If not, a local LLM on your own hardware exists for exactly this reason. Decide before you subscribe to anything.
Infrastructure, then governance, then training. Build the stack, write the rules, train the people. Agents come after all three, not before.
Not sure where you stand?
Book a readiness audit. We tell you what is worth doing, what is not, and what to ignore.
FAQ
AI for an SME in Hong Kong or Australia means replacing work that still lives in WhatsApp, Excel, or paper with systems and agents that run on your own infrastructure. It is not a chatbot subscription and not a tool roundup. It is the order that lives in a chat, the spreadsheet only one person understands, and the file only the filing cabinet knows, replaced by systems that run every day.
Three questions. Does critical work live in WhatsApp threads or someone’s chat history? Does the business depend on a spreadsheet only one person understands? Does anything important exist only on paper? If you answered yes to any of them, you are not ready to buy an AI agent. Fix the infrastructure first.
Infrastructure, Governance, Training, in that order. Build the stack and decide cloud versus local first. Then write the rules for how AI and agents act, shaped by how you actually deploy. Then train the people who own each workflow to manage it. Doing it backwards, buying an agent before the stack and the rules exist, is how AI projects fail.
A readiness audit: an honest pass across the business that says what is genuinely worth automating, what is not, and what to ignore. One audit for an Australian manufacturer with around one hundred staff produced three builds: automated quality documentation, a procurement portal that compares supplier quotes, and a sales planning tool. The audit costs a conversation to scope. Buying the wrong agent costs a lot more.
Use free programmes, like Australia’s AI Adopt Centres or Hong Kong’s SME support schemes, for literacy and first experiments. Avoid buying agent SaaS before your work is out of WhatsApp, Excel, and paper. Hire a consultancy like 3 Peat when you want the stack built on your systems, the rules written, and your staff trained to run it. Months, not millions.