Published on the 28/05/2026 | Written by Heather Wright

Confidence gap slows AI progress…

SMEs have made up their minds on AI, but they haven’t quite figured out how to use it without breaking something.

New data from Xero shows 45 percent of Kiwi small businesses see AI as the biggest opportunity since the internet, yet deeper adoption is being held back by a confidence gap driven by concerns around trust, privacy and lack of clear implementation pathways.

“They’re grappling with how and where to use it safely and confidently, how to move past the early adoption and unlock more meaningful value.”

Australian SMEs are hitting the same wall, with trust in AI decision-making remaining the single business barrier to adoption, according to data from the Australian Government National AI Centre. Its latest data shows that SME AI adoption rebounded to 44 percent in February – the strongest result in several months. But many businesses remain unsure how to stay in control once AI is introduced.

Belief high, execution shallow

According to Xero’s data almost two-thirds (61 percent) of Kiwi SMEs are already proactively using AI. But how they’re using it matters more than how many are using it – and the picture here is less impressive, with Xero saying much of the AI use is ‘only at a surface level’.

Most businesses are teaching themselves through experimentation, with 79 percent learning via trial and error and layering AI onto existing processes rather than redesigning workflows.

Bridget Snelling, Xero New Zealand country manager, says most SMEs aren’t questioning whether AI matters. “What they’re grappling with is how and where to use it safely and confidently, how to move past the early stages of adoption and unlock more meaningful value,” Snelling says.

In practice, that means AI is largely being used tactically – to draft emails, summarise documents or automate small tasks perhaps – but not strategically embedded into how the business actually runs.

That disconnect is reinforced by the barriers SMEs say are holding them back. The Xero data identifies a consistent set: Lack of time, concerns around data privacy and trust and – critically – the absence of a clear roadmap for implementation.

Asked what would help them feel more confident in deploying AI, the SMEs were clear: Practical training and workshops (47 percent), access to trusted, vetted tools built specifically for SMEs (46 percent) and real-world case studies showing how others are using AI (43 percent).

On the government side, SMEs were keen to see clear regulations and accountability standards (57 percent) and enforcement of strong data protections and standards (55 percent), with one in three also wanting the government to provide education and training resources.

A similar story for Australia

The confidence gap isn’t unique to New Zealand. Australian data points to the same pattern – growing uptake, but uneven and often ‘superficial’ use.

The National AI Centre figures show 43 percent of Australian SMEs have reported some level of AI adoption in the quarter to February 2026 – down from 45 percent the previous quarter.

The National AI Centre says while that figure might seem modest, among adopters the news is encouraging with a clear shift from surface level experimentation toward deeper integration, with broad adoption – where AI is embedded across multiple parts of the business – reaching its highest level (eight percent) in seven months. As businesses begin to see tangible benefits (Xero’s report points to significant time savings as a key benefit for Kiwi businesses adopting the technology), they tend to deepen their use – but getting to that point is proving a challenge.

Again, the survey identifies three distinct barriers which help explain why more than half of Australian SMEs are yet to meaningfully adopt AI, with trust the leading barrier at 65 percent. Relevance (54 percent) and cost and skills (20 percent) are also holding SMEs back according to the National AI Centre.

Among the non-adopting businesses, 54 percent believe AI isn’t relevant to their business, something the National AI Centre attributes to an absence of visible, relatable examples of what AI actually looks like in practice for a business like theirs.

The relevance gap is particularly pronounced in industries like construction and agriculture, where fewer than 30 percent of businesses are currently adopting AI. In contrast, the health, education and services sectors are leading adoption, with more than half of businesses in those sectors actively using AI.

“The difference isn’t capability, it is context. Businesses need to see themselves in the story of AI adoption,” the National AI Centre says.

Workers are using AI – and worrying about it

At the same time, the workforce is leaning into AI faster than organisations are formalising its use, creating new layers of risk.

Global research from GoTo (which wasn’t focused on SMEs) highlights a growing tension. While AI is boosting productivity, it’s also triggering concerns about overreliance and skills erosion. Half of employees say they already depend too heavily on AI, and 30 percent say they can’t function without it. Meanwhile, 39 percent believe that reliance is weakening their skill sets.

There’s also a sense of pressure with 60 percent of workers reporting feeling expected to use AI to improve productivity, even when they lack formal guidance or training.

The result is a workforce that is moving quickly, but not always confidently, and often without guardrails. That carries real consequences. Nearly one in four IT leaders surveyed in the GoTo report say AI has made mistakes that impacted customers, clients or their company’s bottom line. And 91 percent worry that AI could make a mistake that negatively affects their company.

Taken together, the Xero and GoTo findings point to a shift in the AI adoption conversation. It’s no longer about access to tools or even awareness of their potential. Businesses already have both. Instead the limiting factor is confidence – in the tools, the safeguards around them and in their own ability to apply them effectively.

Xero’s data breaks businesses down into four distinct ‘archetypes’ based on their beliefs influencing their AI adoption: Explorers, sceptics, trailblazers and pragmatists.

The company says for explorers, who are curious, practical and willing to try new tools but are largely experimental and self-guided, layering AI onto existing ways of working rather than rethinking processes, there’s opportunity to start baking AI tools in. “This could mean developing a deeper understanding of your business processes to better target AI deployment or more experimentation to reimagine workflows with an AI-first lens.”

For the sceptics, small mindset shifts could present opportunities.

“Whether it’s talking to other business owners about how they tailor AI implementation, or taking another look at your end-to-end business processes to identify specific opportunities that align to the capabilities on offer, knowledge and guardrails could be key to building confidence.”

For trailblazers it’s a case of avoiding being caught up chasing the newest, hyped up tools, and instead putting some effort towards mastering strategic integration of the existing tools. “The real opportunity isn’t just doing things faster or using as much AI as possible, but rather using a purpose-fit, context-aligned AI stack to realise even greater benefits within your business.”

And for the pragmatist? “View the time spent learning AI not as a ‘cost’ that takes you away from work, but as a ‘capital investment’ that buys back future time. Focusing on optimising the right processes could mean five hours spent today saves five hours every week – a worthy trade-off, indeed.”

From confidence gap to competitive advantage

There’s a throughline across both reports: SMEs, and indeed larger organisations too, aren’t failing to adopt AI because they don’t see the value. They’re struggling because they don’t know yet how to unlock it safely and consistently.

Businesses that close that confidence gap – by gaining the training, governance and clear use cases – will move from fragmented, task-level gains to sustained operational advantage.