CITIC SEC: Management software landing scenes are rich, optimistic about Agent driving management software growth further.

CITIC SEC released a research report stating that it is optimistic about the further growth of management software driven by Agents, with incremental growth expected in tool software and key industry software as well. Agents have autonomous planning, perception, and decision-making abilities for complex tasks, achieving value enhancement from knowledge assistance to manipulation replacement. Salesforce's new digital labor platform Agentforce is leading the enterprise AI platform towards full transformation. Agentforce validates the capability upgrade and commercial value of AI applications under the Agent form, and the market has strong demand for application landing. With a solid foundation of models and data domestically, the scenario for the landing of management software is rich, embracing the vast ocean of B-end Agents, and a turning point in performance is foreseeable. Key opinions of CITIC SEC: Overview of management software: Encountering short-term difficulties after cloud transformation, Agents empower a new paradigm for management software Management software targets B-end demand, and UFida and Kingdee are leading domestic management software companies. Looking back at the leading companies, UFida and Kingdee have undergone three transformations and upgrades in financial management, ERP, and SaaS, capturing industry demand changes and achieving continuous growth. Currently, after completing the cloud transformation, a tiered operational system based on the PaaS platform has been formed for different cloud products for large, medium, and small enterprises. Since 2022, management software companies have faced slowdowns in revenue growth, declining profit margins, and even losses after the cloud transformation, requiring new growth engines. Agents will help management software open up long-term space. Since 2023, management software has actively embraced generative AI and has gradually evolved towards the form of Agents since 2H24. According to the AGI five-level theory proposed by OpenAI CEO Sam Altman, AI product forms are evolving from Chatbots, Reasoners to Agents, with greater automation in human-machine integration. Agents consist of perception, memory, planning, action, and tools, enabling users to complete entire tasks autonomously without the need for manual operation, bringing value enhancement from knowledge assistance to operation replacement. 2025 is expected to be the year of landing for Agents. From a C-end perspective, overseas companies like OpenAI Operator and Anthropic Computer Use, and domestic companies like Zhifupu AutoGLM are expected to serve as technological bases to support the accelerated landing of high-value AI Agents in various fields; from a B-end perspective, Salesforce's introduction of Agentforce provides a new paradigm for AI Agents to empower management software, with its insightful significance. Lessons from others: Agentforce provides automated digital labor, enriching software value through scenarios such as customer service, marketing, and human resources. Salesforce is a global CRM leader, achieving scene-based layout with SaaS+PaaS, powered by data analysis and AI. The company started its AI layout in 2014, accumulating ten years of experience and launching Agentforce, leading the AI Agent era. The AI strategy has helped Salesforce reduce costs and increase efficiency, with a historical high GAAP operating profit margin of 20.0% in 3QFY25, with customers showing the highest willingness to pay for Agentforce among all AI products. In terms of product features, Agentforce is an autonomous and adaptable digital labor force, integrating Agents, data, and CRM functions, ready for use and customizable: 1) Out-of-the-box Agent, a digital employee that can work 24/7, providing over 100 out-of-the-box industry solutions, allowing users to interact in natural language, such as virtual coaches, virtual interviewers, virtual customer service, etc. 2) Customizable user Agents: Agentforce has a No-code feature, allowing users to customize Agents using natural language. In terms of product value: 1) Externally: Agentforce serves as a professional digital labor force, helping customers unleash productivity, increase ROI, and top customers are actively using it; in the customer service scenario, according to CMSWIRE, the cost of human customer service per interaction is currently between $3.0-$6.5, significantly higher than Agentforce's pricing of $1.0-$2.0 per interaction, saving over 50% of costs. 2) Internally: Strategically, the company is migrating its technical infrastructure and product system to the Agentforce platform for a full transformation towards an enterprise AI platform; in terms of performance, Agentforce not only drives sales of other cloud products, promoting the company's "multi-cloud" strategy, but also has a pricing of $2 per interaction, currently using a hybrid billing model, and is expected to gradually transition to a usage-based billing model in the future, which could increase ARPU. Lessons from overseas: Agentforce, based on "reasoning technology" + "data cloud" + "full scene," three core weapons 1) Technological barrier: The Atlas reasoning engine utilizes RAG retrieval enhancement technology, significantly reducing model hallucination, with accuracy increasing from 40%-70% in the test version to 90%-95%. 2) Data barrier: The Data Cloud integrates internal data like PaaS, marketing cloud, Tableau, and external databases like Snowflake, Databricks, Google Big Query, through "zero-copying" to integrate internal and external enterprise data; according to the "Leadership Next" column of Fortune magazine interviewing the company's CEO, Data Cloud have accumulated 230-300PB of customer behavior data (three orders of magnitude more than GPT4's training data). 3) Scene barrier: The company has conducted nearly a hundred product acquisitions since 2004, continually enriching the Customer360 product matrix and supplementing the company's technical capabilities in data analysis + AI. Agents, as the intelligent body switch of full-process product capabilities, are embedded in Customer360 for functional integration. Domestic practice: Establishing a foundation of models and data, management software embraces the vast ocean of B-end Agents.1) Model support: Domestic models support the acceleration of AI application in all fields. In terms of model capabilities, models such as DeepSeek V3 and Bean Pro have been introduced with universal text model capabilities approaching global leading levels, and multimodal capabilities gradually improving. In terms of model prices, the output price of large model APIs in China is generally priced at 10 yuan or less per million tokens, with many priced even lower, at 1 yuan or less, which is one-third or lower than GPT-4o, helping to promote the widespread adoption of Agent. 2) Data utilization: According to iResearch, unstructured data in enterprises accounts for 80%, but the utilization rate is only 30%. Agent improves the efficiency of utilizing unstructured data; management software collaborates with customers to promote the digital construction of enterprises and has private domain data barriers. 3) Scenario value: B-end business models are relatively clear, and Agent is expected to empower AI in more B-end scenarios by enhancing ROI. Management software companies combine Agent capabilities with their own workflow layout, such as ERP systems serving as the foundation of enterprise digital systems, upgrading to an enterprise-level Agent platform with AI. B-end Agents are implemented in various industries, opening up vast opportunities in the long term. Investment strategy: 1) Optimistic about Agent driving the further growth of the management software market, recommend prioritizing management software companies that have launched Agent features and have clear customer promotion plans. 2) Other generic software such as tool software and key industry software are also expected to benefit from incremental opportunities brought by the penetration and landing of intelligent body products like Agentforce. Risk factors: Development of AI core technology falls short of expectations, improper use of AI causes serious social impacts, AI applications expand slower than expected, downstream customers AI and IT spending fall short of expectations, information security risks, increased industry competition, and heightened geopolitical conflicts.