Agentic AI

APAC Is Racing Into Agentic AI—Without Understanding It

Agentic AI is spreading fast across Asia-Pacific, but the workforce meant to wield it is being left behind. New research from Boston Consulting Group finds that 77% of APAC workers say their employers are already experimenting with or deploying autonomous AI agents—yet only 33% say they properly understand the tools, exposing a widening gap between business adoption and employee readiness.

That gap matters because agentic AI goes far beyond generating text, images or recommendations. These systems can update records, select audiences, trigger customer journeys, adjust offers and coordinate tasks across business functions. The upside is real efficiency and personalisation gains—but so is the downside when decisions get made without sufficient oversight.

Most Deployments Are Still Experiments

Adoption headlines can be misleading. While 77% of APAC workers report their companies are experimenting with or deploying AI agents, only 13% say the tools are already integrated into broader workflows, according to BCG. A further 64% say agents are being used experimentally, in pilots or under human supervision—evidence that most businesses remain in the testing stage rather than running fully autonomous systems at scale.

For marketing teams, that experimentation could mean using AI agents to support campaign planning, audience segmentation, content production, customer service and real-time optimisation. But the value of those applications depends on whether businesses can connect agents to reliable data, clear processes and defined responsibilities.

Without this foundation, AI agents may simply accelerate existing problems instead of solving them. — Belinda Lloyd, Customer Success Lead, Amperity

Customer Data Becomes the Control Point

The stakes rise sharply once agentic AI is wired into customer data and live business systems. An agent might identify a high-value customer, recommend a product, update a record or fire off a message—but if the customer's preferences are outdated or their consent status is unclear, that same action can quietly damage the relationship.

The failure modes are concrete: an agent could recommend a product a customer just returned, contact someone who has unsubscribed, or suppress a valuable customer because their identity was never correctly matched across platforms. That is why trusted customer context becomes a critical control. Before an agent acts, a business needs to know who the customer is, what their latest interactions show, what permissions are in place and what the system is actually allowed to do.

Three Questions to Ask Before an Agent Acts

BCG's findings also flag concerns around human oversight, accountability and bias—signs that many organisations are moving ahead of their own education and governance structures. For marketing leaders, three questions should guide any agentic AI deployment:

• What does the AI agent know? Teams need visibility into the customer data, signals and assumptions informing each action.
• What is the agent allowed to do? Clear permissions should determine whether it can recommend, update, contact, approve or execute.
• Can the organisation see and correct the outcome? Audit trails, feedback loops and human intervention are essential when an action is wrong or inappropriate.

Those controls matter most precisely as agents graduate from isolated experiments into customer-facing workflows.

Commerce Is Bracing for the Same Shift

The move toward agentic AI reaches well beyond workplace productivity. Deloitte's report, "The Future of Commerce: Agentic Shopping in Asia Pacific," found that 29% of APAC consumer businesses are already adopting agentic AI, with adoption expected to climb to 76% within two years.

Deloitte frames agentic AI as a force that could reshape customer engagement, retail operations, physical stores and the wider shopping journey—with agents helping consumers discover products, compare options and make purchases, while businesses lean on them for personalisation, pricing, service and operational planning. But the firm cautions that technology alone won't create competitive advantage: businesses will need to strengthen their technology foundations, business models, security, governance and customer trust. For marketers, that means the customer experience may soon involve not only human customers and brand teams, but AI agents acting on behalf of both sides.

Why Marketers Should Pay Attention

The takeaway from both BCG and Deloitte is blunt: adoption should not be confused with readiness. As autonomous systems gain access to customer records, marketing platforms and commercial workflows, the marketers who win won't be the ones automating the most tasks—they'll be the ones who pair speed with control, backing every agent with reliable customer data, clear accountability and outcomes that stay visible, explainable and correctable.

---

About the Research

The figures draw on two studies. Boston Consulting Group's report on APAC workers found that 77% say their businesses are experimenting with or deploying AI agents, only 33% feel they understand the technology, 13% report full workflow integration, and 64% describe experimental, piloted or human-supervised use. Deloitte's "The Future of Commerce: Agentic Shopping in Asia Pacific" reports 29% current adoption among APAC consumer businesses, rising to a projected 76% within two years.

Background: The Author

This analysis is a thought-leadership piece by Belinda Lloyd, Customer Success Lead at Amperity, a company focused on unifying customer data so brands can build the trusted customer context that governs how—and whether—AI agents are allowed to act.