The AI Proficiency Gap Kiwi Firms Keep Missing

Leaders think their AI rollouts are working. The data says otherwise. Here’s what we’re seeing on the ground in New Zealand, and what to do about it.

A year ago, AI proficiency meant something pretty simple; can your people use ChatGPT without leaking client data, and can they write a prompt that doesn’t embarrass anyone? Most NZ businesses have done a decent job getting to that bar. People know what AI is. They know how to summarise a meeting. They’ve stopped pasting sensitive data into free tools.

That was 2025. The bar has moved.

In 2026, proficiency means something much harder; your people using AI every week on real work that actually moves the business. Not reformatting emails. Not summarising documents they were going to skim anyway. Actual leverage; hours saved, quality lifted, decisions made faster.

And we’re just not seeing it. Not at the scale the hype suggests.

97% of the workforce are either not using AI at all; or using it for tasks so basic they save almost no time. The gap between “I use ChatGPT” and “AI is changing how I work” is enormous.

Those numbers come from a survey of 5,000 knowledge workers across the US, UK and Canada, published by Section in January 2026. We’ve run more than 90 AI workshops across Auckland over the last two years with Spark, Business North Harbour, the Waitakere Business Hub and dozens of individual clients; and honestly, the NZ picture is the same. Possibly worse at the smaller end of town.

Here’s what we think every NZ business leader needs to understand heading into the rest of this year.

1. Usage is not adoption. Adoption is not value.

ChatGPT has 900 million monthly users. More than half of Americans say they use AI. More than half of the knowledge workers in the Section study use it at least weekly. If those were our metrics, we’d all be declaring victory.

But when you look at what people actually do with it:

• 70% are “experimenters”; dabbling with basic prompts a few times a week

• 28% are “novices”; they’ve tried it and bounced off

• Less than 3% have genuinely integrated it into their workflows in a way that drives real productivity

A quarter of the workforce say they save zero time with AI. Another 44% save less than four hours a week. That’s not transformation. That’s a very expensive search engine.

If you’re measuring your AI rollout by logins, licence count, or “are people using it,” your dashboard is lying to you.

2. The real problem isn’t prompting. It’s knowing what to use AI for.

Every workshop we run, the same pattern shows up. We spend maybe 20 minutes on prompting basics. Then someone puts their hand up and asks the real question; “okay, but what should I actually use it for in my job?”

That’s the gap. The Section data confirms it: 85% of knowledge workers have no work-related AI use case, or only beginner-level ones. 26% don’t have a work-related use case at all.

The top ten reported AI use cases globally are pretty sobering:

The number one use case on the planet is “replacing Google.” The tenth is task and process automation; actually valuable workflow integration; at 1.6%.

That lines up exactly with what we see locally. People have been given an LLM and told to “have a play.” What they need is a use case library for their role; the five or ten specific tasks in their week where AI genuinely saves hours, with clear examples showing how.

3. Executives think it’s going brilliantly. Nobody else does.

This is the finding that should make every CEO pause.

When Section asked C-suite executives and individual contributors the same questions about their company’s AI rollout, the gap was extraordinary:

Forty-six percentage points is not a communication problem. It’s two different realities. The CEO thinks the strategy is clear and tooling is in place. The people doing the work think neither is true.

When we sit down with NZ business owners, this is almost always where the conversation starts. Leaders have signed up for Copilot or ChatGPT licences, written a policy, maybe run a lunch-and-learn; and ticked AI off the list. Meanwhile the team is using it occasionally, sheepishly, with no real sense of what “good” looks like.

4. Training alone won’t close the gap

Here’s the uncomfortable bit. The Section study showed that employees who’d received formal AI training from their company scored an average of 40 out of 100 on proficiency. Training helps; trained employees are 1.5x more proficient than untrained ones. But it still leaves them a long way short of actually being productive with AI.

The reason is pretty straightforward. Most AI training is still teaching the 2025 curriculum; what is AI, how does it work, how to write a prompt, don’t paste confidential data. That’s necessary. It’s not sufficient.

The next layer, which almost nobody is teaching well yet, is:

• How to spot a workflow in your job that AI could eliminate or compress

• How to chain steps together rather than one-shot prompting

• How to build a repeatable process with AI in the middle of it; not just a one-off clever output

• How to tell when AI is wrong, and when to trust it

• How to work with automation tools like n8n or Copilot Studio so the AI actually does the work instead of you copy-pasting between windows

This is where the real productivity comes from. It’s also where most NZ businesses haven’t started.

5. What this means for New Zealand businesses

The Section data is US-centric and skewed toward 1,000-plus person enterprises. The NZ reality is different in some important ways; and in some ways the same:

What’s different: We’re smaller and more agile. A 30-person NZ business can genuinely move faster than a 3,000-person US enterprise. There’s less politics, shorter decision chains, and the owner is usually close enough to the work to see where the opportunities are.

What’s the same: Most teams are still stuck at the “summarise a meeting” level. Most policies are vague or copy-pasted off the internet. Most managers aren’t modelling daily AI use. Most use cases are single-prompt tasks rather than proper workflow integrations.

What’s worse: We generally don’t have dedicated AI transformation teams, training budgets, or in-house prompt engineers. For the mid-market NZ business with 20 to 200 staff, AI enablement sits on top of an already-full plate; usually the GM, the owner, or whoever’s loudest about technology.

The good news; because NZ businesses are smaller, a genuine AI practice is actually achievable. You don’t need a transformation office. You need a handful of solid use cases per role, a few real automations running in the background, and someone making sure it’s all actually being used.

Six things to do before the end of Q2

If you’re a business leader reading this and thinking “yeah, that’s us”; here’s where we’d start.

1. Stop measuring AI by licence count. Pick two or three metrics that actually indicate value; time saved per person per week, number of workflows automated, use cases live per team. Track them monthly.

2. Build a use case library, by role. Your finance person, your sales person, your operations person; they all have different work. Give each of them five specific, vetted AI use cases that fit what they actually do. Don’t leave them to figure it out alone.

3. Close the awareness gap with your team. If you’re at the top, assume your picture of adoption is more optimistic than reality. Do skip-level conversations specifically about AI. Ask what’s in the way.

4. Invest in workflow thinking, not just prompting. The next wave of value is chained processes and automation, not better one-shot prompts. Get someone in your business; internal or external; who thinks in workflows.

5. Back your individual contributors, not just managers. The data shows ICs are the most left behind. They’re also doing the most repeatable, automatable work. Give them tools, training and explicit permission to experiment.

6. Accept that the bar keeps moving. The gap between “we’re fine with AI” and “AI is genuinely changing how we operate” will keep widening. Build continuous learning into the business rather than running a one-off workshop and calling it done.

The bottom line

The companies getting real value out of AI in 2026 won’t be the ones with the biggest licences or the slickest policy document. They’ll be the ones that took the boring, unglamorous work seriously; figuring out exactly where AI belongs in each role, building workflows that actually run, training people beyond the basics, and measuring the outcome rather than the activity.

That’s not a technology problem. It’s a leadership one.

And it’s genuinely achievable for any NZ business willing to treat this as a real programme rather than a side project.

Want to know where your team actually sits?

iT360 runs a free AI Readiness session for NZ businesses; a 60-minute working conversation where we benchmark where your team is against what we’re seeing across the market, identify the three or four highest-value use cases in your operation, and map out what a 90-day programme could look like.

Get in touch at it360.co.nz or email us at enquiry@it360.co.nz

About the author. Callum Galloway is Chief Sales & Marketing Officer at iT360. He leads iT360’s AI & Automation division and has delivered more than 90 AI workshops across Auckland in partnership with Spark, Business North Harbour, the Waitakere Business Hub and the North Harbour Business Association.

Source data. Global benchmarking figures in this article are drawn from Section’s AI Proficiency Report, January 2026, surveying 5,000 knowledge workers at 1,000+ employee companies across the US, UK and Canada. NZ observations are from iT360’s AI workshop programme and client engagements.