GenAI系列之四十八：Agent如何重构软件生态？

Investment Rating - The report does not explicitly state an investment rating for the industry Core Insights - Agents enable AI to collaborate with humans, transitioning from a "co-pilot" role to a "pilot" role [3] - Agents simplify the customization challenges faced by software companies, reconstructing the "man-day" model [4] - The capabilities of Agents are currently at a stage similar to the transition from GPT-3 to ChatGPT, with significant improvements expected post-2025 [5][35] - The adoption of AI Agents is anticipated to revolutionize software companies' pricing models, moving towards per-use charges and revenue sharing [5] Summary by Sections 1. Algorithm: A Key Step Towards AGI - AI Agents are more suited for complex, continuous tasks compared to traditional LLMs [10] - The relationship between LLMs and Agents is that LLMs serve as the brain for Agents, essential for decision-making [10] 2. Industry: Customization and Commercial Reconstruction - The software industry faces challenges with high customization demands, which suppress profit margins and productivity [53][58] - The average profit margin in the software industry has declined from 13% in 2010 to 1.5% in 2023 [58] - The introduction of Agents is expected to address these customization issues by leveraging non-structured data and adapting to diverse business processes [68][73] 3. Overseas: Agents Have Started to Take Off - Major companies like Microsoft and Salesforce are leading the charge in developing AI Agents [98] - Microsoft has established CoreAI to support the development and deployment of AI applications and Agents [100] 4. Related Targets - Key domestic players in the Agent space include ByteDance's UI-TARS and Zhiyu's AutoGLM, both of which are making strides in AI capabilities [27][32] - The report highlights various companies and their advancements in AI Agent technology, including Microsoft, Salesforce, and SAP [5][6][7]