AI spending soars, but governance and ROI lag behind

The gist
AI budgets are skyrocketing, but most enterprises are flying blind—governance, data quality, and measurable ROI are nowhere near keeping up.
What to know
- Australian firms expect to spend AUD $35.5 million on AI in 2026, yet only 22% feel mostly or fully ready on governance—well below the 33% global average.
- Nearly all businesses plan to maintain or increase AI spend, but just 1 in 4 have a clear framework to link AI to business goals, leaving finance chiefs to manually review outputs for 30+ hours per week.
- Agentic AI is being unleashed without proper guardrails—only 7% of organizations prioritize governance over speed, and over 70% say the pressure is to deploy fast, not safe.
AI Budgets Outpace Value
Soaring AI investments are undermined by poor data quality and rushed agent deployments, with nearly half of organizations reporting costly errors and unchecked outputs.
By late 2026, enterprise AI adoption was high enough that weak returns had become a mainstream operating problem. CFOtech Australia, citing SAP and Oxford Economics research across 2,600-plus business leaders in 13 countries, found spending and use rising even as readiness lagged: Australian organisations expected to spend AUD $35.5 million on AI this year, up from AUD $27.5 million last year, while only 22% said they were mostly or fully ready in AI governance, below the 33% global average.
The same evidence helps explain why higher investment was not reliably translating into value. CFOtech Australia reported that 64% of Australian leaders ranked integrated data systems as the biggest enabler of AI readiness and 51% pointed to data quality, yet nearly three-quarters of organisations reported poor data quality problems; meanwhile, 42% said they were deploying AI agents faster than they could standardise and govern them, 49% said agents had already taken incorrect actions during pilots or deployment, and 54% said employees were accepting AI outputs without enough scrutiny. November data suggested the mismatch was broadening, not narrowing, as enterprise budgets accelerated faster than paying adoption: Campus Technology, citing Menlo Ventures, reported that enterprise AI deployment spending reached a $176.8 billion annualized run rate by September, up from $33 billion in 2025, while the share of enterprises paying for AI rose from 55% to 65%.
Ownership Gaps Stall Progress
Despite universal AI strategies, most companies lack clear accountability for results, leaving finance chiefs overwhelmed and measurable returns elusive.
The pattern across 2026 is not a shortage of AI activity but a shortage of the operating discipline needed to turn that activity into repeatable returns. TechRadar cited Snowflake saying “Nearly all businesses will maintain or increase AI spend over the next 1-2 years,” yet “Only one in four have a clear framework to use AI to realize business goals,” and while 45% of UK organizations reported modest productivity gains, only 23% had achieved them at scale, showing deployment is outrunning the governance, data, and accountability needed for enterprise value; the drag is visible in finance operations too, where an IDC survey commissioned by Sage found “nearly one in five” European finance chiefs devotes more than 30 hours each week to manually reviewing AI-generated results.
As deployments move into production, the missing layer is ownership and control architecture, not enthusiasm. Fortune reported Melissa Valentine saying generative AI is moving “out of its experimental phase and into something CFOs know well: systematic measurement,” while noting that “rigorous accountability would have been premature” two years earlier, but many companies “still have not decided who should own the job of turning that momentum into measurable business value,” with only 2% of surveyed C-suite executives assigning that role to CFOs; KPMG likewise found nearly all firms had an AI strategy, yet only 8% saw clear returns even as 32% were deploying agents and 27% had multiple agents.
Speed Overrules Supervision
As enterprises race to scale AI, oversight lags dangerously behind, leaving most unable to explain failures or guarantee compliance in live deployments.
The clearest pattern in 2026 is not a lack of AI activity but a failure to update oversight at the same pace. Business Wire found that nearly all organisations, 99%, felt pressure to scale AI in customer experience, yet only 38% had a clear governance approach; 59% said they were scaling quickly while just 39% believed compliance was keeping pace, a “gap between speed and control” that was already pushing AI from experimentation into live interactions without dependable supervision or clearly defined accountability. The practical prescription is straightforward: pick one approved tool with the right plan, including SOC 2 Type 2 certification, rather than letting adoption outrun basic controls.
What reliable oversight looks like is well understood, which makes the gap harder to dismiss as temporary. KPMG’s Q4 2025 AI Pulse Survey found 75% of business leaders rank security, compliance, and auditability as the top requirements for agent deployment, and Morgan Stanley’s AI assistant reached 98% adoption only because, as Jeff McMillan put it, “Every use case tested before deployment. Every output logged. Every decision traceable”; by contrast, Business Wire reported that when AI factories fail, six in 10 enterprises still cannot explain why. That contrast underscores the point: oversight is not mysterious, but many enterprises are still scaling faster than they are standardising accountable, certifiable systems.
Data Chaos Hinders Scale
Enterprises are rolling out AI on shaky data foundations, forcing leaders to retrofit governance and fueling shadow AI as employees bypass weak controls.
The bottleneck is not model access but data readiness. Business Wire reported Semarchy’s survey of 1,000 global C-level executives found data management is now the top AI challenge for 51% of leaders, yet 51% are implementing AI without Master Data Management foundations and 38% without data-quality standards; the consequences are already operational, with 22% reporting AI project delays from data quality concerns, 21% operational inefficiencies from unreliable data, and 19% compliance issues tied to data protection regulations.
What makes this a scaling problem rather than a temporary startup mess is that enterprises are still retrofitting standards after deployment has begun. Business Wire said 77% of leaders have now fully integrated AI considerations into data governance policies, up from 50% who prioritized AI ethics and regulation in 2025, while Semarchy CTO Craig Gravina warned that only 7% of CDOs and 18% of CIOs are seen as holding a chief role in AI strategy; just under half, 48%, are investing in DataOps to bridge the gap and deliver reliable high-quality data products.
In regulated environments, weak data controls also push AI use outside approved systems, making dependable production even harder. PR Newswire said TraceGains’ survey of 423 food quality, safety, R&D and governance professionals found only 41% had formal enterprise AI initiatives while informal workforce adoption ran ahead; 25% cited enterprise-grade security and data protection requirements as a top obstacle, alongside accuracy and trustworthiness concerns, and TraceGains warned shadow AI may be filling the gap, echoing Gartner’s finding that over 57% of employees use personal GenAI accounts for work and 33% paste sensitive information into unapproved tools.
Autonomous Agents, Unchecked Risks
Agentic AI is multiplying inside organizations faster than control systems can keep up, leaving most firms unable to track actions, assign blame, or ensure safe outcomes.
Agentic AI raises the stakes because enterprises are handing software more autonomy before they can reliably constrain or observe what it does. Gartner warned in The Register that, without governance, agents will “multiply and run amok,” a risk made concrete by forecasts that the average Global Fortune 500 enterprise will run more than 150,000 AI agents by 2028, up from fewer than 15 today; yet Avalara found only 7% of organizations prioritize governance over speed while 71% say deployment pressure is focused primarily on speed, and 30% have not updated internal controls for agents taking or recommending actions.
The control gap is not abstract: many firms cannot clearly explain what their agents did, who owns the consequences, or how the systems actually work. Avalara found 44% were only somewhat confident they could explain an AI agent’s actions to an auditor or regulator, 23% said accountability for a significant agent error would be unclear or sit with no one, 16% thought blame would fall to the executive who approved the investment, and 76% lacked in-house expertise to understand their agents; in Australia, 59% cited speed pressure versus just 12% prioritizing governance over pace.






