How do I get started with AI in my small business?

How do I start using AI in a small business?

Pick one task, not a tool. List the jobs that repeat every week, score each on volume, repetitiveness and cost of error, and automate the highest scorer first. As at October 2026, On Autopilot, a Melbourne-based AI consultancy, builds a first automation for $497 one-off, with productised agents live in 5–14 working days. Run it alongside your current process, measure it against a baseline, and only then add the next.

Which task should I automate first?

The one that scores highest on three questions, each rated 1 to 5. Volume: how often does it happen? Repetitiveness: does it follow the same pattern every time? Cost of error: what does it cost when it is done late, badly or not at all today? Multiply the three scores. Multiplying matters: a task that is rare or unpredictable scores low however painful it feels, so it cannot win on cost alone.

Example scoring (an illustration for a hypothetical six-person trades business):

Task	Volume	Repetitiveness	Cost of error	Score
After-hours enquiries and booking requests	5	4	5	100
Following up quotes that got no reply	4	5	4	80
Chasing overdue invoices	3	5	4	60
Drafting monthly social posts	2	3	1	6
Annual insurance renewal paperwork	1	2	3	6

The top two are close, and both sit next to revenue, so either is a good first build. The bottom two are the classic trap: easy to automate, satisfying to finish, and barely noticeable on the bottom line.

A high cost-of-error score means a task is worth doing well, not that the AI should act alone. Keep a person approving anything regulated, irreversible or involving money leaving the account.

How does it change by business size?

The method is the same at every size; what changes is who scores the tasks and how many you take on at once. Sole traders usually start with whatever slips while they are working, small teams with one person’s repetitive load, and larger teams with one workflow per team.

• Sole trader: you are the bottleneck, so score what slips while you are on a job. A paid ChatGPT or Claude plan covers writing tasks; a $497 Quick Start or an AI Front Desk covers the enquiries you miss.
• 2–10 staff: score the list with the person who does the work. They know the edge cases, and they become the approver during shadow mode.
• 10–50 staff: score per team and build one workflow at a time, usually under managed AI. If a role changes significantly, award consultation duties apply; see will AI replace my employees.

How does On Autopilot build it?

In six steps, with a working system usually two to four weeks after the first call. The build itself takes 5–14 working days depending on the package. The rest is a baseline before and a shadow period after, which is what makes the result measurable and safe to leave running.

1. Free 30-minute audit (day 0). Jenn Yang scores your tasks with you and tells you which one is worth automating, what it would cost and the likely payback, or that it is not worth it yet.
2. Pick one workflow and record a baseline (weeks 1–2). Write down the before-numbers: hours on the task, unanswered enquiries and response time. This runs while we build.
3. Build (5–14 working days). We connect to the tools you already use, such as Xero, Shopify, Cliniko, ServiceM8, HubSpot and Slack, load your services, prices and rules, and set the escalation triggers.
4. Shadow mode (about a week). The agent works alongside your current process: it drafts, a person approves, and nothing reaches a customer unchecked. We tune whatever it gets wrong.
5. Go live. The agent handles the routine volume and hands anything urgent, unusual or regulated to a person.
6. Measure at 30 and 90 days. Compare against the baseline, then keep it, tune it or switch it off. Then score the next task.

What does On Autopilot charge to get started?

As at October 2026 (AUD, ex GST), the audit is free. A Quick Start, one automation built and handed over, is $497. Productised agents cost $497 to $2,000 to set up plus $99 to $499 a month; the AI Front Desk is $1,500 plus $199 a month. Managed AI, where we run and keep improving the agents, starts at $1,500 a month. Retainers are month to month and you own the code, prompts and settings.

Every package is on the pricing page. The AI Front Desk suits businesses whose top score is after-hours enquiries, and managed AI suits teams with several workflows. Start with the free audit. If you are still choosing between doing it yourself, a freelancer or an agency, read who can set up AI for a small business.

What should I avoid when starting out?

Buying tools before choosing a task, automating a broken process, and pasting personal information into consumer AI tools. The OAIC recommends, as best practice, not entering personal information, particularly sensitive information, into publicly available generative AI tools. Keep regulated sign-offs with a person too: AI can prepare figures for your bookkeeper or BAS agent to check, but only registered tax and BAS agents can provide those services for a fee.

If client data is part of the task, read can I use ChatGPT for confidential client information first. And before spending on anything, check the payback with whether AI automation is worth it for your business.

Sources

• Guidance on privacy and commercially available AI products, OAIC (checked 03/10/2026)
• Tax Practitioners Board, on who can provide tax agent and BAS services for a fee (checked 03/10/2026)
• Claude plans and pricing, Anthropic, USD (checked 03/10/2026)
• ChatGPT plans, OpenAI’s published USD prices; confirm the AUD amount at checkout (checked 03/10/2026)