Every second headline says AI will transform your business. Meanwhile, you've got invoices to chase, a ute in for service and a staff member off sick — and no time to work out which part of the AI story is real. Here's the short version, from people who implement this for a living: some of it is genuinely, boringly useful, most of it is noise, and the difference comes down to picking the right first project.

The rule: aim AI at the boring stuff

The AI wins in small business aren't glamorous. They're the tasks that are frequent, repetitive and text-heavy — reading, retyping, summarising, drafting, finding. If a task makes your staff sigh, it's a candidate. If a task requires judgement your customers pay you for, keep the human on it.

Three first projects that reliably pay off

1. Document processing: stop retyping paperwork

Supplier invoices, delivery dockets, paper forms, licence applications — someone in your business is reading them and typing the contents into a system. Modern AI reads them instead: it extracts the supplier, amounts, dates and line items, and pushes them where they need to go, flagging anything it's unsure about for a human to check. Businesses processing even 50 documents a week typically save hours, and error rates drop because the machine never gets bored at 4pm.

2. A knowledge assistant: answers from your own documents

Most businesses have their knowledge scattered across policy documents, old emails and one long-serving employee's head. An AI assistant trained on your documents — procedures, price lists, product specs, safety data sheets — lets any staff member ask "what's our process for X?" and get the right answer with a reference, instantly. It's like giving every new hire a patient mentor who has read everything and never takes leave.

3. Email triage: first drafts, not final answers

AI can sort your inbox by what's actually being asked, draft replies to routine enquiries, and route the tricky ones to the right person. The human stays in charge — reviewing and sending — but the blank-page work is gone. For businesses drowning in quote requests and booking emails, this alone can win back an hour a day.

What to avoid (for now)

  • Fully automated customer-facing chatbots with no human fallback. When they're wrong, they're wrong to your customers.
  • AI making unsupervised decisions about money, safety or people. Keep a human approving anything with consequences.
  • Buying "AI-powered" everything. If a vendor can't tell you exactly which task their AI does and how it fails, walk away.
  • Feeding sensitive data into free consumer tools. Free tools have terms of service you wouldn't sign if you read them. Use business-grade services with proper data agreements.

The data question everyone should ask

Before any AI tool touches your business data, get plain-English answers to three questions: Where does the data go? Who can see it? Is it used to train someone else's model? Set up properly — under enterprise terms, in your own tenancy where possible — AI tools can meet the same privacy standard as the rest of your systems. Set up carelessly, they're a leak waiting to happen. This is most of why professional setup matters more than which AI brand you pick.

What a sensible first step looks like

Don't start with a strategy document. Pick one workflow where the numbers are obvious — "we spend six hours a week retyping invoices" — and automate that. Measure the saving. If it works (it usually does), expand to the next workflow with the confidence of a proven result. A first project like this is typically days of work, not months, and we quote it in writing before anything starts.

And if AI genuinely isn't the right fix for your problem? A good implementer will say so — often a simple automation or an integration gets you the same saving with less machinery. That honesty is the difference between buying a solution and buying a subscription.