An AI strategy for a small business starts with one job
Most AI strategy advice is written for governments and enterprises. For a small business it starts smaller: one job, one decision, then the next.
Search for “AI strategy” in Australia and the first page is government AI plans, a national strategy, university short courses and analyst reports. Most of it is written for organisations with an AI budget line, a steering committee and a year to plan.
A small business needs something much shorter: a list of the jobs that eat your week, and an honest answer on each one. That is the whole strategy. Here is how to put it together.
1. Start with the jobs, not the technology
Leave the tools aside for now. Write down the manual jobs in your business, and for each one:
- who does it,
- how often,
- roughly how long it takes,
- and what it costs when it gets missed or done late.
Missed phone calls, enquiries copied from one system to another, weekly reports assembled by hand, the same questions answered over and over. The list is usually longer than expected, and it is the most useful document you will have.
2. Decide job by job
For every job on the list there are three possible answers, and AI is only one of them:
- AI, when the job needs understanding language: answering a call, reading an enquiry, sorting messages by what they are about.
- A simple automation, when the job is moving the same information from one place to another. A form that creates a job in your system does not need a model; it needs the two tools connected.
- A person, when the job needs judgement, or when getting it wrong is expensive and hard to spot.
Some jobs also turn out not to need doing at all. Knowing what not to build is as valuable as knowing what to build, and it costs nothing.
3. Pick one job to start with
The best first job is repetitive, frequent and costly when it goes wrong. Missed calls are a common one: every call that rings out may be a customer ringing the next business on their list. Two real projects started exactly there, a solar installer and a window tinting company, each with an agent that answers the calls nobody can and passes the details to the team.
Build it into the tools you already use rather than adding another app beside them, and decide up front where the data goes and when a person takes over.
4. Test it on real cases before trusting it
Run it where the risk is lowest first: overflow calls only, or one enquiry form, not the whole business at once. Read what it actually does. Adjust the questions, the handover and the wording until it behaves the way your best team member would.
5. Then decide the next job
With one job working, go back to the list. The next decision is easier because it rests on evidence from your own business rather than on a vendor’s promise. That is what makes this a strategy rather than a shopping list: each step is chosen because the last one worked.
Questions to ask anyone selling you AI
- Where is my data processed? If they cannot tell you, do not put your data through it.
- What happens when it gets something wrong?
- What does it cost to run each month, not just to build?
- Which of my jobs would you not use AI for? A good answer names some.
Where to go from here
You can do the first two steps yourself this week with a notepad. If you would rather have someone go through the list with you, say plainly which jobs are worth it, and build the first one, that is how every AI project at Doppio runs: one job first, then the next.