Bringing AI into a New Zealand business costs anywhere from under $100 per person a month for off-the-shelf tools to $100,000 and up for a company-wide platform. Putti has built custom software and AI in Auckland since 2010. What sets the price is the size of the job you hand over, not the tool you pick.
It's the question we get asked most and the one we can least answer on the spot. Nobody can price AI in the abstract, any more than they can price "some software". But the things that move the number aren't a mystery, and once you know them you can look at a quote and tell whether it stacks up.
Why is there no single price for AI?
Because AI isn't a product. It's an approach, and it covers an enormous range.
Handing your team an AI subscription counts as bringing AI into the business. So does building something that reads every invoice as it arrives, matches it against the purchase order, and only bothers a human with the exceptions. Both are honest answers to "we're doing AI". They differ in cost by a factor of several hundred.
Which is why the useful first question isn't about money. It's what you actually want AI to take off your team's hands. Vague answers get vague numbers.
What are the four levels of AI investment?
Almost every business we quote for sits at one of four levels.
| What it does | Typical range (NZD) | |
|---|---|---|
| Level 1, off-the-shelf tools | Speeds up individual writing and research | Under $100 per person a month |
| Level 2, adding a feature | Puts one piece of AI into a system you already run | $5,000 to $30,000 |
| Level 3, custom solution | Hands over a whole workflow, start to finish | $35,000 to $90,000 |
| Level 4, company-wide platform | Spans several teams and several workflows | $100,000 and up |
Level 1 doesn't change how your business runs. It makes individual people quicker, which is worth having, but the workflow stays exactly where it was.
Level 2 is where things start to shift. AI pulls the data out of documents as they arrive, sorts enquiries and routes them to the right person, or answers questions from your own internal documents rather than the open internet.
Level 3 hands over a whole job. The system moves between several of your tools, applies judgement at each step, and escalates only when it hits something it shouldn't decide alone.
Level 4 brings along the things nobody budgets for on the first pass. Security review, audit trails, permissions, and someone whose job it is to own the thing.
These are starting points rather than quotes. As with our guide to custom software costs in NZ, one scoping conversation usually narrows the range fast.
What drives the price of an AI project?
Two projects that both get described as "adding AI" can come in several times apart. Usually it's one of these.
- How many systems it has to touch. The biggest variable, and the one people underestimate. Connecting safely to Xero, MYOB, a CRM and a warehouse system routinely takes longer than the AI work itself.
- The state of your data. Tidy and in one place is quick. Spread across five spreadsheets and three systems, in three different formats, with a column called "notes2" that three people use differently, and cleaning it up becomes half the project.
- How accurate it has to be. There's a real gap between a task that can run at 90 percent and one that needs 99.9. The second brings validation, exception handling and human checks with it, and those cost more than the model ever will.
- Regulation and security. Customer, health or financial data means privacy design and audit trails. Not optional extras. Line items.
- Where people stay in the loop. Working out what AI is allowed to do alone, and where it has to stop and ask, is process design rather than coding. It takes meetings, and meetings take weeks.
What running costs are not in the first quote?
Five things routinely miss the first number, and all of them keep costing after launch.
Model usage is billed on volume, every month, and it climbs as people use the thing. Hosting has to sit somewhere. Maintenance is unavoidable, because the AI models keep changing underneath you and so do the systems you connected to, and a neglected integration tends to break quietly rather than loudly.
Then there's data clean-up, which is the single most common reason a budget blows out. Check what state your data is in before you commit to anything, not after.
And training. A tool nobody uses costs the full amount and returns nothing.
How quickly does AI pay for itself?
Count the hours you get back.
Say one person spends five hours a week processing invoices. Over 52 weeks that's 260 hours, and costed at $40 an hour it comes to about $10,400 a year. Say the automation costs $30,000 to build, which puts it at the top of Level 2.
| People using the same automation | Saved per year | Payback on $30,000 |
|---|---|---|
| 1 | About $10,400 | About 3 years |
| 3 | About $31,200 | About 1 year |
| 10 | About $104,000 | About 4 months |
Look at the bottom row. Building the automation costs roughly the same whether one person uses it or ten, so the saving multiplies while the build cost sits still. That's the whole game.
So when you pick a first project, go for the task the most people repeat the same way. Not the hardest thing on your list.
Where should you start on a small budget?
Don't buy a company-wide AI strategy. It's slower and dearer than it looks, and you end up committing real money before you know what works in your own business.
Pick one task. The good candidates are boring ones: repetitive, time-hungry, light on judgement, and done the same way by several people. In New Zealand businesses the same handful come up again and again. Quoting. Invoice processing and matching. Data entry off forms. Sorting enquiries and getting them to the right desk. The monthly report someone rebuilds by hand every month.
Build that, measure what it actually saved, and let the number decide the next one.
For choosing the first step, see how to start using AI and is my business ready for AI. If you'd rather add AI to what you already run than replace it, adding AI to existing software covers how that works.
How do AI budgets get wasted?
Three ways, mostly.
The first is buying the technology before picking the problem. The subscriptions renew, the dashboard looks impressive, and the work carries on exactly as it did before.
The second is starting before the data's ready. AI won't tidy up a mess on your behalf. Feed it rubbish and you get rubbish back, faster and with more confidence.
The third is leaving the people out. Build the tool, skip the training, and everyone drifts back to the old way inside a month. That one stings, because the software usually works fine.
The Putti view
We've been building software in Auckland since 2010 and have delivered 100+ projects with a zero-failure record. Spark, Hilton, Wendy's, Fletcher Building and Ritz-Carlton have all worked with us.
AI is a powerful tool, not a magic wand. We build custom AI solutions that plug into real workflows, and we'll say so when something isn't worth building. The cheapest AI project is usually the one we talk you out of.
Tell us which job you want AI to take on and we'll give you an honest range for it. Get in touch