The new 2degrees/Deloitte Productivity Propelled report landed last week with a stat worth paying attention to: NZ businesses using AI earned significantly more than those that didn't. SMEs saw around $400,000 more in revenue during FY25. Large businesses, $59.1 million more.
Those are real numbers from Deloitte Access Economics, not vendor marketing. And they paint a genuinely encouraging picture.
Here's the part that deserves a closer look though.
82% adopted. How many integrated?
The same report finds that 82% of New Zealand businesses are now using AI in some form. That's a remarkable adoption rate. But dig into the detail and a pattern emerges: most of that usage is AI features baked into tools people already had. Autocomplete in their email. Smart suggestions in their CRM. The AI is there, technically speaking. It's just not doing much of the heavy lifting.
As Deloitte's Liza van der Merwe put it: "For many organisations, the biggest gains are not from inventing new technologies, but from using what already exists more effectively."
Or, said plainly: buying AI tools is not the same as getting results from them.
The execution gap is wider here than anywhere
This isn't just a New Zealand observation. A Responsive study of 300+ ANZ organisations, published the same week, found that the gap between AI adoption and AI execution is wider in Australia and New Zealand than in any other region they studied. Only the top 20% of ANZ businesses have actually operationalised their AI. The rest have tools. They just haven't rewired their work around them.
Across the Tasman, KPMG's AI Pulse Q1 2026 reports that demonstrating ROI from AI "continues to be challenging." Only 35% of Australian organisations are even prioritising productivity automation, well below the global average of 42%.
And then there's the global picture. McKinsey's latest survey puts it at 73% of enterprise AI projects failing to deliver projected ROI. Not because the technology doesn't work, but because 77% of those failures are organisational. Wrong team. Wrong process. Right tool, wrong context.
Why adoption stalls
Three things tend to happen:
The tool arrives, the process doesn't change. A business adds an AI feature to their workflow, but the workflow itself stays exactly as it was. The AI sits on top of a manual process like a hat on a dog. Technically present. Not contributing much.
Nobody measures the right things. The Budgetly CFO survey found that 75% of SME finance leaders want AI automation, but only 25% are using AI tools. Part of the reason: they can't demonstrate measurable returns. When the CFO asks "what did this save us?", the answer can't be "it feels faster." It needs to be hours, dollars, or headcount redeployed. If you aren't measuring before you automate, you can't prove the value after.
Integration gets treated as an IT problem. It isn't. It's an operations problem. Connecting your voice agent to your CRM is technical work. Redesigning how your team handles inbound enquiries so the voice agent actually reduces load is the work that produces the $400,000 difference.
What the report is really saying
2degrees CEO Andrew Fairgray was direct about it: "AI itself will not lift productivity. The benefits depend on how effectively organisations adopt it, integrate it, and build the capability to use it well."
That word "integrate" is doing a lot of work in that sentence.
The businesses earning $400K more aren't just using AI. They've changed how they operate. The AI is woven into actual workflows, connected to the systems where decisions happen, and measured against outcomes that matter. It's not a feature they toggled on. It's infrastructure they built.
Where this leaves NZ businesses
New Zealand ranks 63rd out of 67 countries on productivity. Capital productivity is down 1.3%. Labour productivity is down 0.7%. GDP per capita is shrinking. The Deloitte report estimates there's a $46 billion opportunity sitting in improved tech adoption and R&D.
That's not hypothetical. It's a gap between where NZ businesses are and where the data says they could be. Closing it doesn't require buying more tools. It requires connecting the ones you have to the work that actually matters.
At EmbedAI, that's precisely what we do. We build AI voice agents, workflow automations, and custom integrations for NZ businesses. Not as a layer on top of how you already work, but as infrastructure woven into your operations. The kind that shows up in your numbers, not just your tech stack.
If you're in the 82% who've adopted but want to be in the group seeing real results, book a 20-minute call. No pitch deck. Just a conversation about where the hours go and what's worth automating.