Most businesses ask "which AI tool should we get?"

That is the second question.

The first one is whether your business is in a state where a tool can help it, and that is something you can work on this month without spending anything.

Here are the five things worth doing first.

1. Write down the job you want handled, badly

Not a polished process map. A list.

Pick the job that eats your week and write down what actually happens, step by step, including the bits that only live in somebody's head.

You will find two things. Steps nobody can explain, and steps that exist because of a problem that was fixed three years ago.

This is the whole game. AI cannot automate a process you cannot describe, and half the value shows up before any software is involved, when somebody finally writes down what the business does.

If a step makes you wince as you write it, mark it. That is where the money is.

2. Find out where your information actually lives

AI works on what it can reach.

If your pricing lives in a spreadsheet on one laptop, your job history is in a filing cabinet, and your customer notes are in the ute driver's phone, no tool can do much with any of it.

You do not need to fix all of that first. You do need to know which of it is true.

Write the list. Pricing, job history, customer records, supplier details, policies, drawings. Where does each one live, who updates it, and is the current version the one everyone uses.

3. Decide what you will not put into a cloud tool

Every business has information it does not want leaving the building.

Employment matters. Health and safety investigations. Client details covered by a contract. Anything covered by the Privacy Act 2020, which is most customer information you hold.

Decide the boundary before someone in your team makes that decision for you by pasting something into a free tool at ten o'clock at night.

Write it down as one page. What can go into an AI tool, what cannot, and who to ask when it is not obvious.

That page is worth more than most AI training courses.

4. Pick one job, not five

The businesses that get value from AI in their first year do one thing properly.

The ones that get nothing try six things at once, get halfway through each, and conclude that AI is overhyped.

One job. The one that eats the week. Usually quoting for trades and manufacturers, sometimes the CRM nobody updates, sometimes the inbox.

Finish it. Measure the hours it gives back. Then pick the next one.

5. Get your people in the room early

Staff are not afraid of AI in the abstract. They are afraid of finding out about it after the decision.

Bring them in while it is still a question. The people doing a job daily know exactly which parts are stupid and repetitive, which is precisely the list you are trying to build in step one.

They also know which parts genuinely need judgement, and those are the parts you should leave alone.

The best AI rollouts are the ones where the team asked for it.

What this is really about

None of the five is technical.

They are the same things that decide whether any change sticks in a business: knowing your process, knowing where your information is, deciding what matters, focusing on one thing, and taking your people with you.

AI just makes the consequences of skipping them arrive faster.

If you want a straight read on where AI would actually save you time, the AI Opportunity Check is twelve questions and about four minutes. It will tell you if AI is not your bottleneck yet, which is a real answer and a common one.

If you would rather see what the work looks like once the prep is done, AI integration for a small business in New Zealand covers the two kinds of work and what each costs.

And if you want somebody to go through it with you, that is what AI consulting and integration is for. Based in Christchurch, working with businesses NZ-wide.