As technology giants such as NVIDIA, Google, Microsoft, Apple, Meta, and Salesforce start placing their bets on AI agents, 2025 is poised to be the "first year of AI agents," marking a significant shift where AI evolves from being merely a "tool" to an "assistant" and even an "agent."

For small and medium-sized enterprises, constructing AI agents necessitates the integration of numerous complex technical components, including large language model invocation, API orchestration, memory mechanisms, tool execution chains, state management, security controls, and more. To ease the adoption of AI agents for enterprises, the AI Agent Marketplace is emerging as a viable solution.

The AI Agent Marketplace can be viewed as a platform where developers can publish, distribute, integrate, and trade AI agents. Enterprises or users can select and deploy diverse agents as needed to accomplish intricate automation tasks.

Whether in the mobile internet era or the IoT era, becoming a platform enterprise has always been an enticing goal, and the same holds true in the age of AI agents. Once the platform attains a critical mass, the aggregation of agent developers and enterprise users will foster a bilateral network effect akin to the App Store. Platform providers can achieve revenue compounding through subscription fees, plugin distribution commissions, advanced feature licensing, enterprise-level services, and other avenues.

In recent months, numerous companies have openly declared their intentions to launch or have already launched their own AI agent markets within their respective domains of expertise. This article delves into the business model of the AI agent market and highlights the future focus of competition.

Giants Compete for the AI Agent Market - Salesforce

In March 2025, Salesforce unveiled AgentExchange, its AI Agent Marketplace, aiming to provide users with a platform to explore, trial, and purchase AI agents. This enables enterprises to directly acquire verified AI components, accelerating the deployment of intelligent applications without the need for ground-up AI agent development.

Salesforce already boasts AppExchange, an application marketplace with over 13 million app installations. What sets AgentExchange apart? It focuses on offering "skills" and "capability" modules for AI agents rather than traditional applications, positioning itself as a market for "digital labor."

Developers and Salesforce partners can list various modular AI agent components on AgentExchange, enabling enterprises to select suitable agents based on their needs for various tasks. Salesforce has announced that over 200 partners, including Google Cloud, Box, Docusign, and Workday, have joined, offering hundreds of pre-built actions, themes, and templates that have undergone security reviews, assisting enterprises in expediting the adoption of AI agents.

AgentExchange is not a fresh start but an "AI agent version upgrade" by Salesforce, leveraging its mature AppExchange marketplace. Focusing on Salesforce's existing business modules like CRM, Sales Cloud, and Service Cloud, it primarily caters to existing medium and large enterprise customers.

Moveworks

Unlike Salesforce, Moveworks, a native AI enterprise, targets a broader spectrum of enterprise-level users. In April 2025, Moveworks launched its AI Agent Marketplace, offering over 100 pre-built agents that customers can install into their respective systems. Currently, Moveworks' AI agents can be utilized in diverse scenarios, including salary planning, reimbursement management, PTO management, timekeeping, recruitment processes, procurement, customer management, and more.

Earlier agents (i.e., Robotic Process Automation, RPA) required enterprises to meticulously write workflows, enumerating each step and considering all possible variants and scenarios. However, with AI agents, coding these workflows has become more streamlined and can be integrated with data sources to provide richer contextual information and understanding.

Moveworks' agent marketplace differentiates itself by offering deep integration capabilities that support enterprise-level workloads and address the process automation needs of complex organizations. The platform not only aids users in discovering AI agent plugins tailored to their organizational requirements (which can extend agent capabilities) but also allows for installation with just a few clicks in a low-code environment through an intuitive interface.

Siemens

While the AI agent markets launched by Salesforce and Moveworks emphasize general enterprise-level management applications, industry giants like Siemens aim to further consolidate their unique competitive advantages in vertical fields.

At the Automate 2025 global automation conference, Siemens, an industrial giant, announced the launch of new Industrial AI agents expected to boost production efficiency by up to 50% for industrial enterprises. Unlike traditional AI assistants that respond to commands, Siemens' new industrial AI agents feature an autonomous neural hub, with their core "Intelligent Conductor" system capable of flexibly scheduling various professional AI agents to collaborate, akin to experienced craft masters.

Siemens' uniqueness lies in utilizing a comprehensive ecosystem to coordinate these agents, enabling them to not only synergize with other Siemens agents but also integrate with third-party agents, achieving unprecedented interoperability. To accelerate applications and innovations, Siemens plans to create an industrial AI agent market hub on the Xcelerator Marketplace, allowing customers to access not only Siemens' own AI agents but also those developed by third parties.

It is conceivable that this marketplace may evolve into an upgraded "Industrial Internet Platform," gathering the world's top manufacturing intelligence solutions.

Business Model of the AI Agent Market

The AI agent industry chain can be segmented into four levels: "Large Language Model," "Agent Framework Middleware," "AI Agent Construction/Deployment Platform," and "Enterprise/Consumer Application Layer." The AI agent platform serves as a pivotal link connecting these levels.

For large enterprises venturing into the AI agent market, achieving commercial success is an inevitable topic. With the influx of more enterprises, the business model of the AI agent market is continuously evolving. The following are several potential models:

① Agent-as-a-Service

Currently, this is the most direct and easily implementable business path. Platform providers or developers create AI agents with specific functions based on large language models and API integration, offering them to enterprises or individual users through subscription, pay-per-use, or licensing models. Examples include HR agents that automate employee onboarding processes, sales agents that generate forecasts based on historical data, and personal productivity agents that automatically archive emails. This model mirrors traditional SaaS services but focuses on "agents with task execution capabilities" rather than mere information inquiry or interface operations.

In October 2024, Microsoft introduced a new suite of AI tools to enhance internal team efficiency. These tools included 10 "autonomous agents" in Dynamics 365 software, capable of autonomously handling specific tasks. These "agents" analyze users' work data from Microsoft services like Microsoft 365 Graph, record systems, Dataverse, and Fabric, assisting users in completing various tasks based on this data. In May 2025, Microsoft further unveiled numerous new products, including multi-agent systems in Copilot Studio. Microsoft stated that this feature would allow enterprises to connect multiple AI agents to "delegate tasks to each other" and "work collaboratively," spanning multiple systems, teams, and workflows—all these readily available AI tools can be considered as Agent-as-a-Service.

② Value-added Model of Agent Plugins and Capability Expansion

The capabilities of an AI agent heavily depend on its callable "toolchain." Therefore, plugins, tools, and connectors surrounding agents can also serve as value-added sources within the business model. These capabilities can be provided by third-party developers or uniformly managed and commissioned by the platform. Similar to "in-app purchases" in the mobile app ecosystem, users can add advanced capabilities to agents, such as image recognition, data analysis, API scraping, PDF parsing, etc., with these plugins carrying independent billing value.

③ AI Agent Marketplace Platform Operation Mode

The platform operation mode has been well-established since the Internet era—the platform provider is responsible for formulating agent development specifications, security audit mechanisms, user interface standards, and aggregating a vast array of "general or vertical agents" built by third parties. Its business model typically encompasses:

• Platform commissions (revenue sharing with developers)
• Enterprise-level deployment licensing fees
• Advanced agent subscriptions (differentiated pricing based on functions, data access, and execution permissions)
• Value-added services for data analysis and operational reports

This platform model exhibits strong network effects and ecosystem stickiness. Once it successfully aggregates a significant number of high-quality agents and developer communities, the platform will attain a strategic position akin to the App Store or enterprise workflow tools.

Although few enterprises have explicitly stated their intention to build a platform at present, many large enterprises have commenced launching AI agent development tools/suites, quietly "laying the groundwork" for future ecosystem prosperity. For instance, in March this year, OpenAI released new tools designed to aid developers and enterprises in constructing AI agents—automated systems capable of independently completing tasks—utilizing the company's own AI models and frameworks. These tools are part of OpenAI's new Responses API, enabling enterprises to develop custom AI agents that can perform web searches, scan company documents, and browse websites, similar to OpenAI's Operator product.

④ Customized Agent Development and Private Deployment Services

For large enterprise users, standard agents often fall short of meeting specific business logic and security compliance requirements. Thus, an important business direction involves providing "agent development and private deployment services" to help enterprises construct exclusive AI agents and deeply integrate them with their existing systems. This model typically charges on a project basis, annual fee basis, or platform license basis.

Final Thoughts

Overall, the business model of the AI agent market distinctly features "platform + ecosystem" characteristics. The focus of future competition will not solely be on model capabilities but also on who can develop higher-quality agents, plugins, and integration capabilities, and who can nurture a more thriving developer and user ecosystem.

References:

Moveworks joins the AI agent library boom, Sina Finance; Moveworks Unveils AI Agent Marketplace to Enable AI Agent Discovery and Deployment in Minutes, Businesswire; Salesforce launches AI agent skills marketplace with 200+ initial partners, SiliconANGLE; Salesforce launches AgentExchange market, simplifying enterprise AI agent deployment and application, Baijiahao; Microsoft unleashes a major move, leaving OpenAI behind? Building its own AI empire, and enterprise customers are buying it!, Jinshidata; OpenAI launches new tools to help enterprises build AI agents, Sina Tech