Where Your AI Budget Should Go In FY2027

As businesses lock in their plans for the new financial year, AI is finally showing up as a real line item rather than something a curious staff member quietly expenses on a personal card. That is a healthy shift. But a budget is only as good as the thinking behind it, and a lot of AI budgets are being set this year with more optimism than evidence.

The encouraging part is that the spending is starting to pay off. Among Australian businesses already using AI, around 79 per cent report productivity gains, according to the National AI Centre. The danger is that, in the rush, the money lands in the wrong places. Here is where I would put it in FY2027.

1. Budget for outcomes, not experiments

The past two years were about trying things, and that phase is not quite over. But it should no longer be where most of the money goes. By now most businesses have a fair idea of which use cases actually returned something and which were fun demos that quietly died. Fund the first group properly and be honest about the second.

A budget built around a specific, painful problem and a number you can measure will always beat one built around ‘innovation theatre’. If you want a rule of thumb, fund the projects that can already point to hours saved, errors avoided or revenue won, and put a hold on anything that still cannot explain what success would look like. Focus on your Return On Investment (ROI).

2. Put a hard ceiling on your usage costs

This is the one I would not skip. AI is increasingly billed by usage, and usage has a habit of running away from you. Around 85 per cent of organisations misjudge their AI costs by more than 10 per cent, and a quarter are out by half or more. The cautionary tales are real, including one company that reportedly spent half a billion dollars in a single month after failing to set limits on its AI coding tools.

You will not rack up a bill like that, but the lesson scales down to any business. The number of organisations treating AI as an active cost concern roughly doubled in a single year, from 31 to 63 per cent, and they are right to. Give one person clear ownership of the AI bill, set alerts that warn you before you reach a cap rather than after, cap usage by project or by person and review the spend each month like any other metered utility on the books.

3. Set money aside for governance before you need it

Late this year the rules around AI and privacy tighten, with new disclosure obligations for automated decisions taking effect in December and a regulator that now has sharper enforcement powers. A short, plain-English AI usage policy is cheap to write and expensive to go without. Budget the time and the advice for it now, while it is a calm planning task rather than a scramble after something has already gone wrong. The line item itself is small. It pays for a written policy, some plain guidance for staff and a review of which tools are touching sensitive data, and it costs far less than the breach or complaint it helps you avoid.

4. Fund the training, not just the tools

The most common waste I see is a business paying good money for capable software that nobody has been shown how to use. The tools change every few weeks and new use cases appear constantly, so a one-off lunch-and-learn does not cut it. Put a real, recurring line in the budget for ongoing training. It is usually the cheapest part of the stack and the part that decides whether everything else earns its keep.

5. Audit what you are already paying for

Before you add anything, look hard at what you already have. Roughly 30 per cent of software licences in the average organisation go unused, and AI subscriptions are among the easiest things to sign up for and then forget. Every licence sitting idle is money that could be funding training or a tool people actually want. A quick audit before renewal season often pays for the rest of your AI plan on its own.

6. Leave room to change your mind

The last line in the budget should be the one you do not fully commit. AI pricing and the models themselves are moving faster than any annual plan can keep pace with, and the cheapest capable option today may not be the one you would choose in six months. Resist signing a long, multi-year deal that locks in this year’s prices and this year’s favourite. Keep a portion of the budget unallocated so you can move when a better or cheaper option appears, and favour tools you can switch away from over those that quietly trap your data and your workflows. Flexibility is a line item worth funding in its own right.

None of this requires a large consulting budget to work out but it may be worth speaking to an expert if you are not feeling confident about your AI strategy.

The businesses that get real value from AI next year will not be the ones that spend the most. They will be the ones that spend on purpose.