Already tipped to potentially change the way businesses operate, artificial intelligence (AI) agents are expected to also bring a much-needed breath of fresh air into customer service. Data issues, though, will still need to be resolved.
Agentic AI has been forecast to fuel $227 billion in global spending this year, with 97% of Asia-Pacific business IT leaders already having implemented or planning to implement AI agents in the next couple of years. The same momentum is expected in customer service, as brands look to enhance user engagement.
As it is, generative AI (GenAI) in the past year has moved from curiosity and experimentation into actual deployment.
Businesses are now looking closely at how their GenAI initiatives are driving value and returns, Zendesk CTO Adrian McDermott said in an interview via video.
It also has pushed standardisation in the customer service industry, with copilots and GenAI applications built to pull information from multiple systems, and generate answers that support customer service delivery, McDermott said.
In addition, there has been further automation in customer interaction and, increasingly, agentic reasoning, as AI agents each take on different tasks across the customer journey, he said.
For example, an AI agent may manage the conversation with a customer request and identify what needs to be done to carry out the task. It then determines the AI agents that have to be activated to fulfil different components to complete the task. An orchestration agent coordinates and manages this, communicating with the different AI agents as they perform each task. It does this until all components are complete and a final resolution is reached for the customer.
Their ability to contextualise and reason plugs a critical link that automation previously could not effectively support in customer experience, McDermott said.
AI agents now can be used to help find the answer, identify what is broken, and enable customers to feel heard, he added.
Shashank Sharma, Adobe’s Southeast Asia and Korean senior director of digital experience, concurred, noting that agentic AI will drive the next era of AI-driven customer experiences, signalling a fundamental shift from reactive to proactive customer engagement.
“Traditional AI systems required explicit programming for specific scenarios, but agentic AI can reason, learn, and adapt autonomously. This means it can handle complex, multi-step customer journeys without constant human intervention,” Sharma said in an email interview.
Consumers choosing AI over static services
And it appears consumers are taking notice.
In Singapore, 18% say they already use AI agents, according to an Adobe study released this week, which polled 500 respondents in the country. Another 31% expect to tap AI agents in their daily lives within the next year, while 75% will use AI more as the technology evolves to support new services and capabilities.
A similar Adobe study in Australia found that 18% of consumers Down Under are using agentic AI, with another 42% expecting to use it within a year.
In Singapore, most consumers are using AI assistants for routine tasks, such as summarising content, but the fastest-growing areas of use are being guided on where to shop and what to buy. AI agents also are replacing traditional search, the Adobe report revealed.
It found that 37% of Singapore consumers are using AI assistants for online shopping, while 41% use these to aid in their travel and 29% in their banking.
Respondents who are less familiar with agentic AI show interest in using it when they gain more knowledge of the technology. Among these, 91% point to scheduling appointments and ordering as compelling use cases, while 90% highlight the technology’s potential to take action and complete tasks.
“What's particularly powerful is its ability to anticipate customer needs rather than simply responding to them,” Sharma said. “Agentic AI can analyse customer behaviour patterns, predict likely issues, and proactively address them before they become problems.”
For example, if a customer's usage pattern suggests they might exceed their service limits, an agentic system can automatically reach out with upgrade recommendations or usage optimisation tips, he said.
“The breakthrough is in contextual persistence. Agentic AI maintains understanding across all touchpoints and interactions, enabling truly personalised experiences at scale,” he noted. “Paired with GenAI, it can create dynamic content, adjust messaging tone based on customer sentiment, and even generate real-time solutions to unique problems.”
"The appetite for AI agents that can complete tasks, make decisions, and streamline daily activities signals a new era of digital experiences,” Sharma said. “For businesses in this region, this represents both an opportunity and an imperative to reimagine how they engage with customers through AI-powered touchpoints that deliver genuine value and convenience.”
AI also enables businesses to better handle the growing volume of customer interactions, McDermott said, which will only continue to climb as communication channels are added and more services are automated, making them easier to consume.
Amidst this increase in service demands, copilots allow human agents to spend more time on high-value conversations that might have been shortchanged in the past, he said.
Asia-Pacific consumers are ready for AI to resolve issues and improve quality of service, he noted, referencing Zendesk 2025 CX Trends Report, which found that 79% in the region had positive views about the technology.
Some 80% in the region note a clear gap forming between companies that have leveraged AI effectively in customer service and those that have not.
According to the Zendesk study, 82% of Asia-Pacific consumers feel it is important for AI agents to have human-like traits, with 80% describing the "friendliness" of AI agents as crucial for a positive experience.
Some already are choosing AI-powered assistants over static online features, with 49% doing so to schedule appointments and 47% for orders, according to Adobe’s 2025 AI and Digital Trends.
This defies a seemingly general disdain for chatbots and automated customer hotlines, which have gained a bad reputation amongst consumers in past years.
Sharma said this usually stems from a “fundamental misunderstanding” among brands of what customers actually want from automated interactions.
“When chatbots and AI assistants are designed around business efficiency rather than empathy for customers, this can lead to rigid, scripted responses, and limitations in their ability to understand context or nuance,” he said.
Instead, chatbots should be programmed to recognise intent, detect sentiment, and respond in a way that feels considerate, he added.
“The key difference is in the quality of the interaction. When AI can provide genuinely helpful, contextually-aware responses, this can accelerate adoption rapidly,” he noted.
AI agents can plug old chatbot gaps
Earlier generations of chatbots also were badly designed, which led to their bad reputation, McDermott said.
User experience has since improved, as organisations begin building richer AI apps and integrating agentic capabilities, he said.
Cosmetic brand Lush, for instance, was able to resolve 60% of customer calls within the first point of contact, after it rolled out a custom AI agent, he noted.
Dubbed Marvin, the AI agent is able to resolve the company’s repetitive queries, such as sales and discounts, order complaints, and discontinued products. The AI agent also can request customer information and add tags to incoming tickets, so agents have the context they need to resolve issues more quickly.
These helped cut resolution time by 5 minutes per ticket and yielded time savings of 360 agent hours per month, according to Zendesk. Lush also improved service agent and manager productivity by 17% and 30%, respectively, generating more than $434,000 in cost savings a year from avoided headcount.
More organisations are turning to AI in search of similar gains in productivity and operational efficiency.
Some 55% of senior executives see AI freeing up critical resources for strategic initiatives, Sharma said, citing Asian findings from the Adobe 2025 AI and Digital Trends report. Another 53% say they are getting returns in marketing effectiveness from applying GenAI.
And unlike other markets that focus their AI application primarily on chatbots and customer support, organisations in this region are leveraging the technology across the entire customer journey, he noted. This encompasses generating creative content and orchestrating complex, personalised experiences.
Pointing to Changi Airport Group (CAG) as a use case, he said the Singapore transport company is tapping AI-powered insights to create more personalised travel experiences that “anticipate passenger needs before they're even expressed”.
CAG is looking to better leverage its data and apply agentic AI to improve operational efficiency and ease customer journey. It is hoping the technology can recommend steps the company can take in various scenarios.
If a flight is delayed, for example, AI agents can automatically trigger the necessary actions. These agents also can activate cleaners or make timely announcements to passengers, further reducing the need for human intervention.
This can be a “game-changer” in a market where standard operating procedures are the norm, said Jeffrey Loke, CAG’s senior vice president of pricing and commercial strategy.
Sharma said: “We're now in an era where creativity, marketing, and AI are combining to transform customer experience, and enterprises all across the globe are investing in people and technology to elevate these experiences.”
New tech, same data challenges
However, as much as data is deemed valuable in AI adoption, it also remains a major challenge in its deployment.
The Adobe report reveals that 88% of Asian organisations cite fragmented data as a barrier in delivering real-time personalised customer experiences. Only 28% view data as a strategic asset.
“It creates a cascade effect where marketing teams have ambitious personalisation goals, but the underlying data architecture is unable to support real-time decision-making,” Sharma said. “The fundamental issue is that organisations are treating data as a byproduct of business operations rather than a strategic asset.”
He noted that brands often underestimate the compound effect of poor data foundation, which will hinder their ability to effect personalisation and decision-making.
“Yet, many businesses are investing heavily in AI capabilities, while neglecting the underlying data infrastructure that makes those capabilities meaningful,” he said.
He urged organisations to move from reactive data management to proactive data strategy, where leadership teams must regard data integration as a business imperative, not an IT project.
They also should implement unified customer data platforms with robust governance frameworks built from the ground up.
“Rather than trying to solve data quality issues downstream through cleansing processes, organisations should implement quality controls at the point of data collection, with automated governance systems that set up usage alerts to prevent privacy policy violations while enabling teams to market responsibly,” Sharma added.
And with data necessary to power AI, Zendesk believes AI agents need two things to function: rules of the business and the tools to run the business.
This has driven the vendor's focus to help its customers build a Knowledge Graph -- to provide AI with the business rules -- as well as connect and automate workflows across their backend systems, through its Action Builder, McDermott said.
Large language models (LLMs), too, play an important role in powering the knowledge AI agents need to function effectively.
Just as important, however, are models that are built for specific purposes, he said.
McDermott anticipates further innovation in the area of AI reasoning, including post-training work on open source models.
As GenAI and agentic AI technologies continue to advance, he also expects more intelligence to be applied to customer service, driving a redesign of standard operating procedures.
Some of these changes already are taking place, with customers feeling less frustrated from not having to repeat themselves as they look for resolutions, he said.
“I think we’re now seeing the ability [of AI agents] to triage and really understand what the customer wants…and [be able to] do more effectively,” McDermott said.
When customers choose AI over human
So what will it take for customers to seek help from an AI over a human?
Sharma believes the shift is already happening, but it depends on the quality of the interaction.
“Customers will choose AI when it can provide immediate, accurate, and contextually relevant responses, without the friction of wait times or business hours constraints,” he said.
For routine inquiries, customers often prefer the predictability and immediacy of AI. They want to check order status, schedule appointments, or get basic information without having to explain their situation to multiple people, he added.
“Ultimately, customer preference will depend on the complexity of their needs and the emotional stakes involved,” he said. “For straightforward transactions, AI will increasingly become the preferred channel. For complex problems or emotionally charged situations, human interaction will remain essential.”
