New Zealand firms race ahead with AI despite data gaps

New Zealand organisations are pushing ahead with artificial intelligence projects despite weaknesses in data readiness, according to Locus. Research cited by the data integration specialist found 65 per cent of local organisations are already running active AI initiatives.

An Omdia survey of more than 1,100 senior technology and business decision-makers across Asia-Pacific, commissioned by Boomi, found only 39 per cent of New Zealand organisations have a platform-led approach to integration. It also found 34 per cent cannot effectively measure whether their AI initiatives are succeeding, while 43 per cent have no dedicated AI budget.

The figures suggest a mismatch between AI adoption and the work needed to prepare underlying data systems. Many organisations are still relying on disconnected systems and manual processes as they pursue AI programmes, Locus said.

"AI is only as reliable as the data you feed it. Across councils, utilities and transport operators, we still see critical information sitting in disconnected systems and being moved between them by hand. Put an AI model on top of that and the errors simply arrive faster," said John Arnerich, Group Director, Locus.

Shadow systems

The research also highlighted concern over so-called shadow integrations, informal system links created outside central oversight. Some 72 per cent of New Zealand respondents said these unmanaged connections were disrupting data quality and reducing confidence in information.

Although 89 per cent of New Zealand organisations identified data integration, access and governance as a key priority, only 38 per cent said they had AI-specific data governance policies in place. At the same time, 86 per cent expected AI initiatives to sharpen the focus on data quality and governance, and 84 per cent said they were working to reduce tool and technology sprawl.

The pressure is not limited to one part of the economy. Separate research from Experian found 76 per cent of New Zealand financial institutions are using agentic AI to support underwriters, yet 55 per cent said data quality and governance were among the reasons AI deployments fail.

Respondents to the Experian survey also identified poor data quality and fragmented systems as operational barriers. In that research, 41 per cent cited poor data quality and 37 per cent pointed to fragmented data systems.

Manual processes

Locus said these operational issues often show up in everyday workflows rather than in AI systems themselves. It pointed to examples from public sector and infrastructure organisations where data tasks remained manual until integration work was introduced.

At NZ Transport Agency Waka Kotahi, a process involving the download, compression and transfer of photo attachments for newly installed network devices had previously taken a full day. After FME was integrated into the survey workflow, Locus said, the same task now takes about five minutes.

Auckland Council also cut the time needed to handle water resource data requests after automating submissions, approvals and delivery, according to Locus. Turnaround times fell from as long as 10 days to three or four.

Wellington Electricity, which receives daily retailer files in inconsistent formats to maintain network records, now runs an automated pipeline without manual handling, according to the company. Locus presented these examples as evidence that integration work and process automation can address data bottlenecks before organisations expand AI use.

"The organisations getting real value from AI did the unglamorous work first. They connected their systems, automated the manual handling and made sure people could trust where every number came from. That foundation is what allows AI to move beyond a pilot," said Arnerich.

Investment plans

Spending intentions suggest organisations are preparing to put more money into these issues. The Omdia research found 93 per cent of New Zealand organisations plan to increase or maintain spending on data integration, preparation and orchestration over the next 18 to 24 months.

Locus argued that the first step for IT leaders is to review how data moves through their organisations, including spreadsheets, scripts and informal links created outside formal oversight. Recurring manual transfers, inconsistent formats and unclear ownership of key datasets often sit behind broader problems with trust, governance and measurement, it said.

The survey data suggests those issues remain unresolved for many organisations even as AI projects move from pilot stages into active use. For companies trying to assess returns on AI spending, the fact that 34 per cent of respondents cannot measure success effectively stands out as one of the clearest signs that data foundations are still lagging behind adoption.