Asia-Pacific AI agent market to surge 14-fold to US$11.2 billion by 2030, Omdia says

An AI agent is software that uses AI models to plan and carry out several connected steps. It can handle work such as analyzing data, writing a report and operating software. This matters because users can delegate a larger job instead of repeatedly directing every small action.

A chatbot that only responds to individual instructions usually waits for each new prompt. An agent can pursue a task across multiple steps. Tencent’s WorkBuddy, for example, can edit Word, Excel and PowerPoint files, then connect with Notion or GitHub to retrieve information and complete related work.

The difference is autonomy and task continuity. Agents can link actions together and work with existing business systems. Their spread also creates new challenges, including monitoring performance, choosing suitable models and making sure the output delivers business value rather than simply using more processing.

Omdia expects the AI agent software market across Asia, Australia and New Zealand to reach US$11.2 billion in 2030. The regional figure includes mainland China and Hong Kong. It compares with an estimated US$782 million this year, showing an exceptionally rapid expansion in just four years.

The market is projected to grow more than 14-fold. Omdia also expects agentic software’s share of regional AI software revenue to rise from 1.7 per cent this year to 12.3 per cent in 2030. That suggests agents will become a much larger part of the wider AI software business.

China is a major force behind the forecast. Chinese technology companies are deploying agents across office and industrial operations. The State Council has set targets for next-generation intelligent devices and agents to exceed 70 per cent penetration by 2027 and 90 per cent by 2030 under its AI Plus initiative.

AI agents are expected to spread quickly because China has set explicit adoption targets and major technology companies are pushing deployment. The State Council wants next-generation intelligent devices and agents to exceed 70 per cent penetration by 2027 and 90 per cent by 2030 through its AI Plus initiative.

Businesses can use agents for practical work rather than isolated demonstrations. Agents can analyze data, write reports and operate software across office and industrial operations. Alibaba’s AgentCore helps companies build, run and manage agents, while Tencent’s WorkBuddy can edit office files and connect to Notion and GitHub.

Cloud services remain the main way companies monetize AI, and lower-cost open-source models are accelerating enterprise adoption in China. Chinese firms are also using open-source models to attract developers and users. Still, deployment requires substantial engineering expertise, and intense price competition may limit how much cloud providers can invest in infrastructure.

Chinese technology companies are building platforms that make AI agents easier for businesses to deploy. These systems matter because companies need more than an AI model: they need tools to create agents, connect them to existing systems, monitor their performance and manage their operation.

Alibaba unveiled AgentCore, an enterprise platform designed to help businesses build, run and manage AI agents. It includes performance monitoring and connections to existing business systems. Tencent is expanding WorkBuddy, an agent workspace that can edit Word documents, Excel spreadsheets and PowerPoint presentations directly.

WorkBuddy can also connect with Notion and GitHub to retrieve information and carry out tasks. Together, these examples show how agents are being placed inside ordinary workflows. The companies are accelerating deployment to capture adoption in China, while open-source models are helping attract developers and users. Deployment still demands substantial engineering expertise.

AI agents could make competition among cloud providers more complicated. An agent-based application may depend on an integrated combination of an AI model, software and computing infrastructure. If those parts are closely connected, customers may find it harder to move their applications to another provider.

The key mechanism is migration difficulty. Moving an agent to a different cloud could require transferring not only the application and its model, but also the optimisations built around them. Those optimisations may be important to how the agent performs its tasks, so changing providers could involve substantial technical work.

This could make customers harder to win away and strengthen the position of providers with integrated platforms. The article does not say that switching will become impossible. It says migration could become more complicated as cloud providers combine models, software and infrastructure in agent-based services.

Tokens are units of data processed by AI models and are commonly used to charge customers. Using more tokens can increase the amount of processing, but it does not automatically improve the usefulness of an answer or the business result. That makes token consumption an incomplete measure of success.

Omdia chief analyst Su Lian Jye said companies should assess the value of AI output and improve efficiency by matching tasks with appropriate models. A smaller or more suitable model may handle a task effectively, while a larger model could add cost without creating equivalent value.

This issue matters as enterprises adopt agents for many kinds of work. Companies need to connect AI spending with useful outcomes, not simply maximize activity or processing volume. Omdia’s warning places return on investment at the center of adoption, alongside engineering demands and the need to manage agents inside existing business operations.

An AI agent uses an AI model to plan and carry out a multi-step task. The model supplies the reasoning and task planning, while software systems provide access to documents, services and business processes. Computing infrastructure supplies the resources needed to run the model and application.

For example, Tencent’s WorkBuddy can edit Word, Excel and PowerPoint files and connect to Notion and GitHub. It can retrieve information from those services and then carry out related tasks. This shows the agent moving between a model, software tools and workplace content rather than producing an isolated reply.

The connection among these layers is becoming strategically important. Omdia said integrating models, software and computing infrastructure could make customers harder to move between clouds. The article also notes that deploying open-source models requires substantial engineering expertise, so making the pieces work reliably remains a significant business challenge.