Australian firms boost AI spend despite weak returns

Elastic has published research showing that many Australian businesses are increasing AI spending despite doing little to measure financial returns. The survey covered more than 500 senior AI decision-makers at organisations with at least 50 employees.

The findings point to a widening gap between AI adoption and proof of business value. One in three businesses exceeded their AI budget in the last financial year, and half plan to increase spending again over the next 12 months.

At the same time, only 8% of respondents said they track AI against revenue growth, cost savings or productivity gains. Instead, businesses are more likely to monitor prompt volumes, token consumption and general usage than commercial outcomes.

That mismatch is already affecting projects. Nearly a third of organisations said they had paused, cancelled or scaled back AI deployments because the return did not justify the cost, while another 28% said they were reviewing deployments for the same reason.

The results suggest finance leaders and boards are asking harder questions about whether AI spending can be justified. They also indicate that organisations are moving beyond early experimentation towards tighter scrutiny of budgets, governance and measurable results.

"AI has moved very quickly from experimentation to a serious budget line item in Australia," said Jeremy Pell, Country Manager ANZ, Elastic. "Most businesses are still only measuring AI usage, but we predict that in the new financial year, this is when that changes. Leaders will want to see real value, not just usage reports. The businesses that come out ahead will be the ones that can control spend, audit risk and show what the investment actually delivered."

Budget pressure

Spending plans are also affecting other parts of corporate technology budgets. Among those reallocating funds to support higher AI investment, 16% said the money would come from IT infrastructure and operations, 12% from existing software licensing, 10% from headcount or hiring budgets, and 8% from cybersecurity and information security.

This suggests AI is no longer funded only as a separate innovation project. It is beginning to compete directly with core operational and security spending.

Elastic warned the shift could create fresh risks, particularly where security budgets are reduced as AI use expands. It linked that concern to a broader increase in the complexity and exposure created by new AI systems.

"The findings also show AI investment decisions are increasingly affecting broader technology priorities across businesses. With AI increasing the risk of security threats, it is concerning to see even a small number of businesses redirecting cybersecurity spend. This is not an area businesses should be deprioritising," Pell said.

Data readiness

The survey also identified data quality as a central issue behind poor AI results. When AI tools underperformed, 32% of respondents blamed poor data quality, compared with 14% who pointed to limitations in the AI models themselves.

Yet relatively few organisations had treated data readiness as a formal requirement before deployment. Only 28% said they had formally assessed whether their data was ready before rolling out AI tools, while 17% carried out only a light-touch review and 8% said they deployed without any formal data quality assessment.

These figures suggest many organisations are trying to scale AI before establishing reliable data controls. In practice, that can make it harder to explain costs, trace errors and show whether a system is improving business performance.

"You cannot build a reliable AI business on data you do not trust," said Pell. "The businesses that get real value from AI will be the ones that do the unglamorous work first: understanding where their data lives, whether it is accurate, and whether it is fit for purpose. If your retrieval layer is feeding the model junk, you are not just paying to compute, you are paying to compute junk data. AI ROI is an observability problem before it is a budget problem."

Governance gaps

Governance appears to be under strain as businesses expand into more autonomous AI systems. Only 31% of respondents said they had a centralised view of how many AI agents or autonomous workflows were running across their organisation.

Almost half, or 47%, said they were moderately or extremely concerned that AI adoption was outpacing their ability to govern it. The most common safeguard was human approval before high-impact actions, cited by 33% of businesses.

More systematic controls were far less common. Only 13% said they had usage logging and monitoring in place, 11% conducted regular risk reviews or audits, and just 2% had a formal incident response process for AI-related failures.

Weak oversight raises questions about how well companies could respond if an AI system caused operational or compliance problems. Only 22% said they were very confident they could identify what had gone wrong and why.

Despite those concerns, 50% of respondents said they planned to expand their use of AI agents. That leaves many organisations trying to increase deployment while basic visibility and accountability measures remain incomplete.

Workforce effects

The research also pointed to changes in how work is being redistributed inside businesses. Three-quarters of respondents said staff were using time saved from automation for higher-value work such as strategy, product development, customer engagement and upskilling.

Another 45% said they expected to create new AI-specific roles that did not previously exist in their organisations, and 18% said they were already hiring for them. This suggests AI investment is affecting workforce planning as well as technology budgets.

"Every business in Australia is being asked the same question right now: what did we actually get for what we spent on AI, and is it worth spending more? The businesses that can answer that honestly, because they have built the data foundation and the visibility to know, are the ones that will pull ahead. This year does not have to be a bigger version of last year. It can be a smarter one," Pell said.