Avanade survey: most ANZ firms can't show AI has paid its way

What the research found

On 4 November 2026, TechDay published the Australia and New Zealand results from Avanade's latest AI research. Avanade is a global technology consultancy, and its figures don't flatter the region. Of the ANZ respondents, 68% said AI hasn't yet produced financial returns that justify what they've spent on it. Another 66% said nobody has formally calculated the return from AI and had the number reviewed by finance or senior leadership. Only 43% use ROI or business value targets to track AI progress. Even so, 88% said they need to speed up AI transformation urgently.

The regional numbers come from a global study called Build Beyond Limits: The AI Reality Report, which Avanade released from Seattle on 3 November. Vanson Bourne ran the fieldwork in August and September 2026. The sample was 3,600 business decision-makers in 11 markets, all at organisations with more than US$500 million in global annual revenue. Across the full sample, 77% described their organisation as established, advanced or AI-first. Only 18% said their teams had fundamentally changed how they operate around AI, and 75% admitted their organisation overstates its AI maturity in public.

The ANZ results look much the same. According to the TechDay report, 84% of local respondents haven't reshaped how they work around AI. Just 23% reported a transformational improvement in employee productivity, and the same share reported one in operational efficiency and costs. One figure stood out against the global result: 39% of ANZ respondents believe colleagues are using AI in ways their compliance team wouldn't approve, compared with 32% globally. Ben Beath, Avanade's chief AI officer, said: "Too many organisations are measuring AI progress by activity rather than impact."

Who was surveyed

Every organisation in the sample earns more than US$500 million a year, so the survey describes large enterprises. Most New Zealand businesses are a small fraction of that size. Avanade also sells AI transformation services, and its recommendation to redesign workflows around AI matches what it sells. Both points are reasons to treat the percentages as indicative.

Other recent surveys point the same way, though. On 28 October, IT Brief reported results from an Omdia survey commissioned by Boomi of more than 1,100 senior decision-makers across Asia-Pacific. In its New Zealand figures, 34% of organisations can't effectively measure whether their AI initiatives are working, 43% have no dedicated AI budget, and only 38% have AI-specific data governance policies. In Australia, CPA Australia's 2026 Business Technology Report, covered on 19 October, found that 88% of businesses use AI in some form but only 10% have embedded it deeply. Those three studies used different samples and methods, and all three found the same gap between using AI and being able to show what it's worth.

Why NZ firms should care about the measurement gap

Measurement matters most at renewal time. When an AI tool or project comes up for another year of funding, an organisation with no baseline can only offer anecdotes. In a tight budget round, finance teams tend to cut spending nobody can quantify, whether or not the project was actually helping. Many New Zealand mid-sized firms run AI pilots on modest budgets with sponsors who also have other jobs, and those pilots are the easiest to lose under that kind of pressure.

The shadow AI figure makes measurement harder. If four in ten respondents think colleagues are using AI outside approved channels, IT and finance can see only part of the organisation's AI use. Any return calculation built on that partial picture will miss both costs and benefits, along with the privacy and data-handling risks that come with unapproved tools.

The link between the 84% figure and the 68% figure is worth noting too. Personal AI assistants often save individual staff time, but saved minutes spread across a team rarely show up as lower costs or higher revenue unless someone changes the process around them. An invoice approval process that still needs four human handoffs costs about the same after each person drafts emails a bit faster. The firms in the Avanade data that reported transformational gains are the minority that changed how work moves through the business. Spreading licences more widely didn't get them there.

What to do next

New Zealand organisations don't need a large programme to close the evidence gap. They do need discipline on a few live use cases before they commit more money. A practical sequence:

• Write down a baseline now. For each AI use case already running, record the current cycle time, cost per transaction, error or rework rate and staff hours. A use case without a baseline has nothing to be measured against.
• Agree the method with finance first. Avanade's finding that 66% of ANZ firms haven't had AI returns reviewed by finance or leadership suggests most organisations skip this step. Get the CFO or finance manager to sign off how benefits will be counted before the review date.
• Count time saved and money saved separately. Freed-up hours become a financial benefit only when they cut overtime, avoid a hire or go to revenue-generating work. Report the two figures apart so nobody double counts.
• Find out what's actually in use. Run a short anonymous survey or review network and expense data for AI tools, then publish an approved list with clear rules on what data can go into each tool.
• Redesign one process from end to end. Pick a high-volume workflow such as supplier invoice processing, service desk triage or quote preparation, and rebuild the steps around what the AI does well. A single redesigned process with measured results makes a stronger case than ten scattered pilots.
• Set a review date and an exit rule. Decide in advance what result would justify scaling the use case, and what result would end it.

For firms working out where AI spending should go and how to measure it, DataForge's AI consultancy page sets out how we approach use case selection and governance. Whatever approach a firm takes, the Avanade results suggest the first useful step for most NZ boards is to ask for a single AI return figure that finance has checked.