What should my business use AI for first?

'Okay... so what should I use it for first?'

Someone asks me this in nearly every workshop, once Claude is open on their laptop and the novelty of getting it to write a limerick about their accountant has worn off. It's a fair question, because the demos make AI look like it can do anything, and 'anything' is a terrible starting point when you run a business with eleven jobs on the go before lunch.

My answer is the same every time. Start with the work that's boring, full of back-and-forth, or done in big volume, because that's where AI is reliable right now and where most of your hours leak out without you noticing. Keep judgement, values and relationships for yourself. And start from a problem you already have (the shiny tool can wait).

Why should I start with the problem instead of the tool?

Starting from the problem stops you building something clever that saves nobody any time. In the room I put it like this: 'Instead of thinking about "What can Claude do for me", I want you to flip your thinking to "What am I trying to accomplish?" Then find where AI fits.'

My favourite version is the hardware store one. Buying a power drill doesn't mean you suddenly need to find things to drill. Plenty of businesses meet AI exactly that way, with a new subscription and a vague sense they should be drilling something.

You can see it in the numbers. The National AI Centre's SME AI Pulse for December 2025 to February 2026 found that 54% of small and medium businesses not using AI said it isn't relevant to them, and 19% of SMEs said they simply don't know how to use it in their business. I'd bet a good chunk of both groups looked at the tool first, saw chatbots and image generators, and decided (reasonably!) that none of it had much to do with fencing a paddock or running a physio clinic. Start from the paperwork around the paddock and the picture changes fast.

A full room of people working on laptops during an AI workshop Jess Tresidder ran at Stone & Chalk in Adelaide
A full room, laptops open. People build during the session rather than watch slides.

What are the 3 Bs of AI use cases?

The 3 Bs are the quickest filter I know for spotting a good first task: Boring, Back-and-forth, Big volume (and data heavy!). Run any task through all three. A task that hits two or three of them is a strong first candidate for AI, and a task that hits none is probably yours to keep.

Boring is the work you could do half asleep, like copying details from an enquiry form into a spreadsheet or reformatting the same document for the fourth time this month. Back-and-forth is anything that bounces, such as the email that needs a reply, then a redraft, then a 'just checking in', or the meeting time that takes six messages to settle. Big volume means the same task done dozens of times, usually with data in it, like fifty supplier invoices a month or two hundred customer reviews you've never had time to read.

Notice what's missing. None of the 3 Bs involve relationships, critical thinking or creativity (unless you count updating the CRM as relationship work, which I'd gently argue against).

Tip the week in and let the 3 Bs do the sorting.

Here's how a fairly normal small business week sorts itself out.

TaskBoringBack-and-forthBig volumeWhere to start
Answering the same customer questions by emailYesYesYesAI drafts from your FAQs, you send
Turning meeting notes into actionsYesSometimesYesTry AI first
Keying in supplier invoicesYesNoYesTry AI on one supplier first
Pulling a monthly report from three spreadsheetsYesNoYesTry AI first
Finding a meeting time with a clientYesYesSometimesA booking link before anything fancier
Writing a quote for a new clientSometimesYesNoAI drafts the wording, you set the price
Calling a long-time customer about a mistakeNoYesNoKeep it human
Deciding who to hireNoNoNoKeep it human

The rows with two or three yeses are where I'd start. The National AI Centre's own list of ways to use AI opens with drafting everyday emails, and it adds a rule I'd underline twice: a person always checks and edits before sending.

How do I know if a task is worth handing to AI?

The SIGH test is the gut-feel version of the 3 Bs, and it works on how a task makes you feel. Look for work that makes you:

  • Start over repeatedly
  • Involves information overload
  • Gets done again and again
  • Has you thinking "ugh..."

The last one is the best data you've got. That little sigh when a certain email hits your inbox is your brain flagging a task it has done too many times, and repetition is the part AI handles well. Keep a note on your phone for one week and jot down every task that gets a sigh out of you. By Friday you'll have a better list of use cases than any vendor demo will give you.

Fair warning, the fix won't always be AI. I once showed a director at one of the big consulting firms Control-K, the shortcut that turns highlighted text into a link, after he'd spent years pasting full URLs into his emails. He was thrilled, and I was a little bit horrified on his behalf (all those sighs, and the fix was two keys).

What should I never hand over to AI?

Keep the work where judgement, values and relationships do the heavy lifting. When I ask a room what they wouldn't want to outsource, the answers end up in the same places every time: judgement, values, relationships, accountability, taste, context and decisions.

'A personal trainer can help you exercise... but if they lift every weight for you, you haven't actually got stronger. Don't outsource the muscle you're trying to build.'

So the phone call to the customer you let down is yours. So is deciding who to hire, what to charge the client who always haggles, and whether to take the job at all. AI can help you think any of those through (it makes a decent sparring partner at 9pm when nobody else is awake), and the call and the name on it stay with you.

The same National AI Centre survey found that around 65% of SMEs not using AI pointed to distrust in AI decision-making or a strong preference to keep human control of their processes. I think that instinct is a good one, and it fits neatly with starting small. Keep the decisions, hand over the data entry, and you never give up control in the first place.

What should I do with the time AI frees up?

Spend it on the human work you've been squeezing in around the admin. The question I end most sessions on is this one: 'What could I do 10x more of if the boring 70% disappeared?'

For a tradie, that might be getting quotes out the same afternoon as the site visit, which beats writing them on Sunday night. For a small accounting practice, it might be one more proper conversation with each client a year, which is the part they'd happily pay more for.

Jobs and Skills Australia's Gen AI Capacity Study found generative AI is more likely to augment human work than replace it, and that matches what I see in rooms. Plenty of people walk out of a session with a list of hours they didn't know they were losing, and a mild argument with their colleagues about what to spend them on.

Once you've got a shortlist, the next question is which tasks deserve a proper automation, and that needs a different filter. I've written up how I score them in how to find AI automation opportunities in your business. Our Untangle workshops are built for sorting a team's week together, laptops open, and the aim is that you don't need me afterwards.

Start with whichever task made you sigh this morning (you know the one).

Questions people ask

Is it safe to put customer information into an AI tool?

Use the South Road test: would you be comfortable seeing it on a billboard on South Road tomorrow? If not, keep it out of free or consumer tools until you've checked the data settings and your organisation's policy. My golden rule is simple. If the AI doesn't need the personal information to do the job, don't give it the personal information.

My business is hands-on. Is AI relevant to a trade or a farm?

Mostly through the paperwork around the work. AI can draft the quote, chase the invoice and turn a voice note recorded in the ute into a job sheet, while the fence and the hot water system stay yours. The 3 Bs still apply, so look for the admin you do at night after the real work is done.

Should I train my team or automate something first?

For most teams, training comes first, because a workshop is the cheapest way to find out what's worth automating. Once people have used AI on their own tasks for a few weeks, they can name the job eating their week, and that's the point where building an automation makes sense.

Which AI tool should a small business start with?

Whichever mainstream assistant your team can already get to, such as Claude, ChatGPT or Microsoft Copilot. You don't need to know every tool. Models, interfaces and names will change, and the useful questions stay the same: what can it do, where does it fail, what information does it need, and what should you trust it with?

Sources

  1. National AI Centre, AI adoption insights: December 2025 to February 2026
  2. National AI Centre, Explore ways to use AI
  3. Jobs and Skills Australia, Australia's AI Transition (Gen AI Capacity Study)