How APAC businesses are turning AI pilots into productivity gains

Competitive pressure is accelerating AI adoption across APAC, but lasting productivity gains depend on disciplined execution and strong governance.

Key takeaways:

AI is moving from pilots to everyday operations, with value and productivity gains now the priority for middle-market businesses.

Scaling AI usage requires more than technology, calling for the right data, skills, and cost controls.

Strong governance and human oversight are business critical as AI becomes more widely used throughout organisations.

Across Asia Pacific (APAC), middle-market businesses are moving from individual AI experiments to wider exploration and embedding the technology in everyday operations. The challenge now is turning competitive pressure into sustained productivity gains, proving where AI use cases add real value, and implementing the right governance around it to support scale.

“Businesses do not want to be left behind, but they are still evaluating where AI can deliver greater economies of scale and how people and AI can work together. More value will emerge when AI is integrated into business processes rather than used alongside them as a separate experiment.”
Mithun Shetty
Partner
India

Pressure is building faster than preparedness

Desire to adopt and innovate with AI is strong, but readiness remains uneven across the APAC region. In April 2026, the Asian Development Outlook’s AI Preparedness Index found a clear gap between APAC’s more advanced economies and many developing markets, likely due to differences in computing capacity, connectivity, and data infrastructure.

Despite this, competitive pressure and expected value are driving adoption. Dr. Suresh Surana, Founder of RSM in India, said businesses are responding to both forces: “The gains are very significant in terms of quality, coverage, and speed for existing applications as well as innovative applications which AI has made possible. Clients and stakeholders are expecting AI-embedded solutions and competition is intensifying the heat. The focus is now on ‘AI in use’ with ‘guardrails in place’ and sharing specific productivity improvement outcomes rather than treating AI as a meeting room discussion.”

Elsewhere, wider economic conditions are influencing the pace of AI adoption. In Australia, for instance, diminished productivity growth is leading businesses to find new sources of efficiency. The Productivity Commission reported that multifactor productivity, a measure of how efficiently businesses combine labour, capital, and technology to produce output, declined by 0.5% in 2024–25, remaining below the country’s long-term average.

Mathavan Parameswaran, National Technology Leader at RSM in Australia, said businesses are “looking to AI to improve productivity, conscious that their competitors are doing the same. They are certainly feeling the pressure to invest in AI and deliver on productivity gains whilst providing better client service.”

Where AI has the greatest potential to add value

The use cases progressing fastest tend to involve high volumes of code, documents, transactions, or customer interactions. Rajesh Maniar, Partner at RSM in India, identified marketing, IT, and research as three functions that could see significant productivity benefits from AI across industries. Regarding use cases beyond office-based assistance, Maniar also pointed to quality control in manufacturing and retailing.

Cybersecurity is seeing a similar pattern with AI being applied to repetitive, data-heavy work. Glenn William S. Alcala, Chief Information Security Officer and Partner at RSM in the Philippines, said: “Cybersecurity is adjacent to software engineering, so a lot of security operations work is being automated, almost like intelligent robotic process automation. We are also seeing more use cases emerge in R&D and security research.”

However, he cautioned that some processes require human oversight: “The manual aspects, particularly in security testing, cannot be automated away with AI based on what we are seeing today.” Making this distinction is crucial for businesses wanting to scale up their use of AI. While automation can create time efficiencies, expert judgement remains necessary, especially when dealing with evolving security threats or where the cost of error is high.

Meanwhile in the professional services space, Caroline Li, Tax Director at RSM in China, shared that: “AI tools have become part of daily work for both tax advisers and clients. Businesses are using AI agents to find answers and analyse tax information, meaning that firms need to add value beyond traditional tax advisory work. At RSM China, we are responding by building our own agents to scan tax risks and support tax due diligence, saving time and cost.” This indicates a shift in the role of advisers towards review, interpretation, and more complex work.

Even an AI powerhouse must prove value case by case

China stands out as APAC’s AI powerhouse, combining strength in research, frontier-model development, industrial deployment, and state-backed implementation. Stanford University’s 2026 AI Index reports that China leads globally in AI publication volume, citations, and patent grants, while the performance gap between leading Chinese and US models has effectively closed.

The country’s national “AI Plus” programme also pairs ambitious adoption targets with extensive governance measures, giving China exceptional capacity to develop and apply AI across many industries. Even with these advantages, however, enterprise adoption is neither automatic nor uniform.

In July 2026, China’s Ministry of Industry and Information Technology published a shortlist of 285 typical AI application cases spanning energy and mineral resources, steel, automotive manufacturing, education, embodied intelligence, and large-model applications. Yet a Q1 2026 survey of over 2,000 industrial enterprises by the Cheung Kong Graduate School of Business (CKGSB) found AI penetration was approximately 10%.

Among businesses that had not adopted AI, 79% said the technology was not applicable to their operations. Sally Yu, Partner at RSM in China, suggests the contrast between national capability and enterprise adoption is clearest in manufacturing: “Companies will only invest in AI for production when they see clear productivity opportunities. In complex industrial settings, data may be unavailable or inconsistent. Error tolerance is low and processes differ between plants.” A successful model in one production stage may therefore be difficult to reuse elsewhere.

There is a broad pipeline of industrial applications, but many businesses still require evidence that AI can improve production processes before committing capital at scale.

“Proving AI’s value in isolated use cases is relatively easy but scaling it across organisations remains difficult.”
Sally Yu
Partner
China

AI’s pilot-to-scale divide

Scaling AI successfully requires a clear business case, the right skill sets, and cost management that ensures investment does not erode the value created.

“What prevents AI pilots from producing measurable gains is often a lack of focus on why the project was started, followed by poor execution.”
Mathavan Parameswaran
National Technology Leader
Australia

An AI pilot can be built around a bounded dataset and a small group of motivated users. Deployment across a business, meanwhile, requires common data definitions, integration with existing systems, clear ownership, and a way to measure outcomes. Crucially, the initiative must also remain anchored to a clearly defined business problem.

What’s more, businesses need to have the skills to use AI effectively at scale. The International Monetary Fund estimates around half of jobs in Asia Pacific’s advanced economies are exposed to AI, with exposure meaning that AI is likely to change or support some of the tasks within those roles.

By comparison, around one-quarter of jobs in emerging and developing economies face similar exposure. As adoption expands, productivity gains will rely on employees having the skills to use AI effectively, interpret its outputs, and apply judgement where needed. Workforce planning and skills development therefore become part of the scaling challenge.

Alongside strategic focus and workforce capability, cost can also become a constraint as AI moves from experimentation to wider use. Greater scale brings greater consumption, making it important to understand how and where resources are being used before costs begin to outpace the value being created.

“The cost of AI usage can be prohibitively high. A model may keep processing without producing corresponding value or desired outcomes, while the costs continue to accumulate in the background.”
Rajesh Maniar
Partner
India

That makes cost management another important part of scaling responsibly. Greater visibility into usage and a clearer link between the resources consumed and the value created can help prevent experimentation from becoming unnecessarily expensive.

The governance challenge of AI at scale

As AI moves from isolated use cases into wider business operations, governance becomes critical to protecting data, establishing accountability, and defining where human oversight is required.

Shetty identified “the need for revalidation of data sources” as a leading concern, with human oversight being critical. The challenge is not only that models can produce incorrect outputs, but that organisations need clear processes for detecting errors and assigning responsibility. That is why governance must be built into AI deployment from the outset.

Regarding data security, Surana said organisations need “a proper framework (such as ISO 42001 and 27701), standard operating procedures, and rigorous implementation to prevent confidential information from entering public domain and for ensuring responsible use of AI.” Australia is moving in the same direction with Parameswaran adding “sound risk management with privacy controls and accountability is increasingly expected before deployment because cyber incidents and regulatory failures can quickly become reputational issues.”

“Privacy and confidentiality remain key barriers to moving AI into production. End-to-end consulting and audit engagements will also require more involvement from partners, senior professionals, and subject-matter experts, not less.”
Glenn William S. Alcala
CISO & Partner
Philippines

Final takeaways

For middle-market businesses, the emerging model is one in which AI handles repeatable, data-heavy activity while people retain responsibility for judgement, exceptions, and relationships. Parameswaran framed the real productivity opportunity as “not just putting AI tools in employees’ hands, but redesigning workflows.”

That is the common thread across APAC’s varied markets. Competitive pressure may start the conversation around AI and individual tools may prove the concept, but lasting productivity gains will depend on whether organisations can connect technology to a disciplined operating model that combines infrastructure, cost and security governance, and human capability.

“The next question is how you apply AI across different teams, work units, and at an overall organisational level. That will take thoughtful planning, roadmapping, implementation, and a change in mindset because you must get team members and clients on board and aligned around common processes.”
Dr. Suresh Surana
Founder
India