CICC: AI+ office opens the next trillion-yuan market; the B-end may become the long-term focus of commercialization.

CICC released a research report stating that AI office is an important deployment point for Agents moving toward general productivity and is expected to become a market with mid-term potential for large-model applications. On the supply side of AI office, improved model capabilities, better training infrastructure, falling costs, and market education are accelerating the penetration of office Agents. Currently, domestic office Agents enter by acquiring C-end customers, with profitability not being a short-term goal. The bank believes that the B-end has higher willingness to pay and greater depth, and may become a key commercialization market in the long term. CICC's main views are as follows: Market space: Demand for office Agents is rising steeply, and the long-term market space may reach US$1 trillion On the demand side, since 2026, demand growth for knowledge-based work in Codex and Claude Code has exceeded that for development-type tasks, and products such as Microsoft 365 Copilot have also preliminarily validated willingness to pay; the bank believes that although office ARPU is lower than Coding, a user base nearly 20 times larger is expected to support considerable market space. On the supply side, improved model capabilities, better training infrastructure, falling costs, and market education are accelerating the penetration of office Agents. Core elements: Office Agents have different competitive priorities across five core dimensions 1) Models: Beyond the intelligence frontier, greater emphasis is placed on capabilities such as computer operation, multimodality, and tool calling, as well as cost efficiency and speed. 2) Harness & Context: The importance of accumulating and managing context and workflows becomes more prominent. 3) Customer entry points: Relying on existing customer entry points helps reduce customer acquisition and deployment costs. 4) Products: Greater emphasis is placed on ease of use, and vendors often lower the usage threshold through UI interfaces and preset template skill libraries. 5) Ecosystem: Internal and external ecosystem collaboration helps expand the boundaries of Agent capabilities. Industry landscape: Multiple types of vendors are accelerating competition, and long-term commercialization may mainly be carried by the B-end Model vendors use frontier model capabilities to promote scenario generalization through third-party ecosystem cooperation; internet vendors have relatively complete endowments in models, cloud, C-end traffic, and B-end customers, and the domestic "battle for entry points" is extending further from Chatbot to Agent; traditional office software vendors control delivery entry points, and customer resource advantages help reduce customer acquisition and deployment costs. Currently, domestic office Agents enter by acquiring C-end customers, with profitability not being a short-term goal. The bank believes that the B-end has higher willingness to pay and greater depth, and may become a key commercialization market in the long term. Risks Agent generalization falls short of expectations; inference cost decline falls short of expectations; security and compliance risks