AI investment accelerates, but cost and ROI concerns remain

The organisations that succeed will be those that can deploy AI safely, securely and consistently, with the right foundations in place to turn investment into tangible business value.

What is driving businesses across APAC to accelerate AI adoption despite ongoing cost and performance concerns?

Wong: The pace of AI adoption across APAC reflects an existential inflection point for businesses, alongside strong ambition to harness AI’s potential. Many organisations are no longer looking at AI as a future-facing innovation project. They are treating it as a capability they need now, whether to improve productivity, drive efficiency, enhance customer experience or keep pace with competitors.

Expereo’s latest Enterprise Horizons 2026 APAC survey shows this momentum is visible across the region, though at different levels of intensity.

AI or machine learning is a priority for 61.3% of organisations across APAC, rising to 72.0% in Australia, 70.0% in Thailand, and 68.0% in Malaysia. Singapore sits at 54.0%, broadly in line with Indonesia and Vietnam at 52.0%, showing that AI investment has moved firmly onto the enterprise agenda across APAC, even as markets move at different speeds.

However, the same data also shows that AI investment momentum is not always matched by careful evaluation or ROI analysis. Across APAC, 36.8% of organisations say they are investing aggressively in AI with little evaluation, driven by fear of being left behind, while 41.2% are investing because they see AI’s potential, sometimes without careful evaluation or ROI analysis.

This discipline gap is visible across markets: Australia and Vietnam show some of the strongest signs of aggressive AI investment with limited evaluation, at 45.5% and 44.4% respectively, while Malaysia stands out for investment driven by AI’s perceived potential but without careful ROI analysis, at 57.5%.

AI adoption in APAC is being accelerated by both opportunity and pressure. The opportunity is clear, but the next phase will require greater clarity around business outcomes, ROI and execution.  Businesses need to move beyond adoption for adoption’s sake and focus on scaling AI effectively, with the right governance, skills, infrastructure and operational readiness in place.

With many organisations investing in AI before establishing clear ROI frameworks, how does this impact business outcomes?

Wong: The biggest risk is that organisations confuse AI activity with AI impact. A business can be running multiple pilots, deploying new tools and increasing AI budgets, but still struggle to prove whether those investments are improving performance, customer experience or ROI in a measurable way.

Our survey underlines this point. While 40.0% of organisations across APAC say their AI implementations have exceeded or significantly exceeded expectations in delivering outcomes or ROI, the majority have not reached that level of performance.

Among APAC organisations where AI has fallen short or only partially met expectations in delivering outcomes or ROI, three issues stand out: cost and ROI, data quality, and AI performance. More than half cite costs being higher than expected or ROI not being achieved, at 54.3%, followed by inadequate or poor-quality training data at 48.6%, and AI not delivering the expected performance or outcomes at 45.7%.

Across markets, the reasons AI implementations have fallen short or only partially met expectations in delivering outcomes or ROI vary. Malaysia and Indonesia more frequently cite costs being higher than expected or ROI not being achieved, at 80.0% and 64.3% respectively. Thailand and Australia point more sharply to inadequate or poor-quality training data, at 71.4% and 66.7% respectively, while 62.5% of organisations in Vietnam cite AI not performing as expected. This reinforces that organisations need to address different barriers before they scale AI further – from cost and ROI visibility, to training data quality, to whether the technology is delivering the expected performance and outcomes.

This is why ROI clarity matters. AI should not be treated as an isolated technology investment, but as a business transformation programme with defined outcomes, clear ownership and ongoing performance measurement. The survey also shows that 40.7% of APAC organisations are concerned about losing track of AI-related costs or ROI once AI is embedded throughout the business — reinforcing that ROI visibility remains a regional challenge as AI scales.

So, the issue is not speed versus caution. It is whether organisations have the visibility and discipline to scale AI responsibly. With the right data, governance, cost controls and connectivity in place, AI has a much stronger chance of becoming a measurable driver of business performance.

Why is network and infrastructure readiness becoming increasingly important as AI deployments scale?

Wong: Network and infrastructure readiness matter because AI performance is only as strong as the network it runs on. As organisations move from pilots to wider deployment, AI needs to draw on data across systems, clouds, teams and markets. If connectivity is inconsistent, resilience is weak or cloud access is not optimised, the impact affects AI performance, operating costs, service reliability and users’ confidence in the technology.

Many organisations are not fully ready for this shift. Across APAC, only 9.0% of organisations say their network infrastructure is fully ready to support new AI, cloud and digital initiatives, while 30.0% say it can support current use but will soon need to be upgraded or replaced. The readiness gap is uneven across the region: Indonesia stands out, with no respondents saying their networks are fully ready, while 48.0% say their network infrastructure will soon need to be upgraded or replaced.

The need for stronger infrastructure is also reflected in what organisations say they need next. Across APAC, 57.0% say they need more flexible and scalable networks, while 46.0% say they need to improve resilience or reliability to maximise network uptime. Across markets, however, infrastructure needs differ. Thailand shows particularly strong demand for flexibility and scalability at 74.0%, while Indonesia and Malaysia place greater emphasis on cloud connectivity optimisation at 48.0% and 46.0% respectively.

This is why connectivity should not be treated as a back-end IT issue. It is part of the AI operating model. It affects whether AI can scale across markets, whether performance is consistent, whether costs are visible, and whether businesses can rely on AI in day-to-day operations. For APAC enterprises, infrastructure readiness will increasingly determine whether AI remains a promising pilot or becomes a dependable business capability.

What should organisations take into consideration as AI moves from experimentation and pilots to real-world business adoption?

Wong: As AI moves into real-world adoption, organisations need to ask a different set of questions. It is no longer only about whether the technology works. It is about whether the business is ready to govern it, secure it, measure it and scale it consistently.

Governance becomes critical at this stage. AI systems are increasingly influencing decisions, workflows and customer interactions, so enterprises need clear controls around how data is used, how outputs are reviewed and where accountability sits. This is becoming more urgent as AI risk moves up the boardroom agenda.

Across APAC, 54.3% of organisations identify the creation of new security risks as a significant future barrier from AI use, while 39.3% are concerned about reputational, financial or legal harm from anomalous AI output or behavior. Thailand and Indonesia show heightened concern around new security risks, at 66.0% and 60.0% respectively, while Malaysia stands out on reputational, financial or legal harm from anomalous AI behavior, at 52.0%. Operational readiness is just as important. AI cannot scale on model performance alone. It needs reliable data, internal skills, resilient infrastructure, cloud connectivity and visibility across performance and cost, especially in a region where enterprises operate across markets with different levels of digital and network maturity.

For leadership teams, the practical takeaway is simple: moving AI from experimentation to real-world adoption requires organisations to address the hurdles that can limit its effectiveness, from governance and security to data, skills and infrastructure. 

The organisations that succeed will be those that can deploy AI safely, securely and consistently, with the right foundations in place to turn investment into tangible business value. Underpinning this is the network. Without the resilience, scalability and cloud connectivity needed to support AI at scale, businesses will struggle to turn AI ambition into measurable outcomes.