经过近两年高速演进，生成式人工智能正从"技术可行"走向"价值可行"的关键验证期。围绕人工智能 +如何落地，行业已形成若干共识，但在更关键的路径选择上仍存在深刻分歧。这些分歧的走向，将决 定AI能否真正成为新质生产力。 三个共识 （一）落地瓶颈已从供给侧转向需求侧 过去两年，"人工智能+"的讨论重点主要集中在算力、模型和数据等供给侧要素。进入2026年，这些问 题虽未完全解决，但已不再构成规模化落地的唯一核心矛盾。真正制约AI扩展应用的，正在转向需求 侧：真实业务需求是否成立，组织是否愿意为AI调整流程，以及AI带来的效率增量能否覆盖引入与改 造成本。 麦肯锡2025年调研显示，在受访的中大型企业中，88%已在至少一个业务职能中使用AI，但仅三分之一 实现规模化部署；62%的企业在试验AI智能体，仅23%完成规模化落地[数据来源：麦肯锡《2025年人 工智能全球调查》（Th e Stat e of AI: Global Survey 2025）]。目标不明确、集成准备不足、难以证明商 业价值位列阻碍因素前三，均指向需求侧。 （二）企业级AI落地面临定制化困局 数据显示，当前AI解决方案约70%需要定制，仅30% ...

刘琼 腾讯研究院 经过近两年高速演进，生成式人工智能正从"技术可行"走向"价值可行"的关键验证期。围绕人工智能 +如何落地，行业已形成若干共识，但在更关键的路径选择上仍存在深刻分歧。这些分歧的走向，将决 定AI能否真正成为新质生产力。 三个共识 （一）落地瓶颈已从供给侧 转向需求侧 过去两年，"人工智能+"的讨论重点主要集中在算力、模型和数据等供给侧要素。进入2026年，这些问 题虽未完全解决，但已不再构成规模化落地的唯一核心矛盾。真正制约AI扩展应用的，正在转向需求 侧：真实业务需求是否成立，组织是否愿意为AI调整流程，以及AI带来的效率增量能否覆盖引入与改造 成本。 麦肯锡2025年调研显示，在受访的中大型企业中，88%已在至少一个业务职能中使用AI，但仅三分之一 实现规模化部署；62%的企业在试验AI智能体，仅23%完成规模化落地[数据来源：麦肯锡《2025年人工 智能全球调查》 （Th e Stat e of AI: Global Survey 2025） ]。目标不明确、集成准备不足、难以证明商业 价值位列阻碍因素前三，均指向需求侧。 （二）企业级AI落地面 临定制化困局 数据显示，当前AI解决方案 ...

Core Insights - The SuperCLUE-VLM multimodal visual language benchmark for December has been released, with Google's Gemini-3-pro scoring 83.64, leading the rankings, while ByteDance's Doubao model scored 73.15, showcasing the competitiveness of domestic models [1] Group 1: Benchmark Results - The evaluation assessed multimodal large models across three dimensions: basic cognition, visual reasoning, and visual applications [1] - Gemini-3.0 and GPT-5.2 have achieved generational leaps in multimodal understanding and autonomous collaboration capabilities [1] Group 2: Market Trends - The domestic and international large model iterations have entered a new phase characterized by "deep reasoning + agents," with AI applications entering a dual-driven period of performance realization and edge explosion [1] - Doubao's daily usage has surged to become the third highest globally, indicating strong market adoption [1] Group 3: Investment Insights - According to a report from China Merchants Securities, the investment logic in the AI industry chain is shifting from "computing power competition" to "application value" [1] - There is a significant focus on AI-driven software and high-growth edge hardware companies as key investment areas [1] - The ARR of B-end software, exemplified by Salesforce Agentforce, has increased by 330% year-on-year, marking a substantial commercialization phase for AI agents [1]

Core Insights - The Federal Reserve has lowered interest rates for the third consecutive time, indicating easing financial conditions that may lead to increased mortgage demand, especially in regions showing signs of a housing rebound [1] - The mortgage market is facing challenges due to outdated technology in financial institutions, which may hinder their ability to meet rising demand and improve profit margins [2] - There is a growing interest in agentic AI within the mortgage industry, which can streamline processes and enhance efficiency [3][4] Group 1: Challenges in the Mortgage Industry - Many banks and lending institutions still rely on legacy technology that is not equipped to handle increased mortgage demand, leading to inefficiencies [2] - A Freddie Mac study revealed that the average cost for lenders to originate a mortgage was over $11,500 as of summer [2] Group 2: Innovations through Agentic AI - Agentic AI can automate routine tasks in the mortgage process, significantly reducing the time required for loan processing and underwriting, thus lowering origination costs [8] - Proactive risk management is enhanced by agentic AI, which can perform automated underwriting and sophisticated risk modeling to identify potential issues early in the lending process [9] - AI-driven automated valuation models (AVMs) are transforming property appraisal by analyzing vast amounts of data quickly and accurately [10] Group 3: Customer Engagement and Relationship Building - Agentic AI enables intelligent indexing, allowing lenders to create a comprehensive customer experience by aggregating various data points [13][14] - This technology can facilitate cross-selling and upselling opportunities, enhancing customer relationships and potentially recommending additional financial products [15][16] Group 4: Regulatory Considerations - The use of AI in lending must address potential biases and ensure explainability in decision-making processes to comply with regulations [17][18] - Lenders are encouraged to maintain human oversight in critical decisions to balance AI efficiency with the need for judgment and empathy [18] Group 5: Future of the Mortgage Industry - The mortgage lending industry has the potential to become a leading example of effective AI implementation, creating a more efficient and predictive ecosystem [19] - The combination of agentic AI technology and skilled human oversight presents a transformative opportunity for forward-thinking lending institutions [20]

Core Insights - Workday is strategically investing in generative AI while ensuring a structured governance framework for evaluating these investments [3][4] - The partnership between IT and finance is crucial for maximizing the effectiveness of AI initiatives, especially given the high failure rate of AI pilots in the industry [6][8] - Workday is open to vendor consolidation and replacement based on the performance and cost-effectiveness of AI tools [7] Investment Strategy - Workday has established a monthly review process involving IT and finance to assess the feasibility and impact of various AI tools [4] - The company evaluates generative AI use cases that have been in production for at least six months to measure their performance against key indicators like productivity and revenue generation [4] Leadership and Accountability - The executive leadership team meets bimonthly to align on the overall AI strategy and to discuss updates on the generative AI roadmap [5] - Transparency in the investment process is emphasized to ensure accountability for AI-related decisions [5] Market Dynamics - Research indicates that many organizations using generative AI are experiencing limited returns on their investments, highlighting the need for careful pilot testing [6] - Workday pilots new AI features with short contracts to mitigate risks associated with unproven technologies [6]

Core Insights - Varonis Systems, Inc. has launched AI identity protection for Salesforce Agentforce, enhancing visibility and control over data accessed by AI agents [1][2] - The new features allow organizations to assess and manage AI agent risks while ensuring the security of sensitive data [1][2] Product Features - The capabilities include automatic discovery and cataloging of AI agents, providing a unified view of data sensitivity, permissions, and activity [9] - Varonis enables organizations to enforce least privilege access by analyzing prompts and responses for sensitive data exposure and policy violations [9] Market Position - Varonis is recognized as a leader in data security, focusing on protecting data across various environments including SaaS, IaaS, and hybrid cloud [6][7] - The company offers a range of security solutions such as data security posture management, data classification, and identity protection [7]

Core Insights - Haleon plc, a leading global consumer company specializing in everyday health, has partnered with Salesforce to enhance engagement with pharmacies and healthcare professionals globally [1] - The collaboration will utilize Salesforce Life Sciences Cloud for Customer Engagement, Data Cloud, and Agentforce to support Haleon's global sales force of 4,500 [1] Company Overview - Haleon plc focuses on consumer health products and aims to improve its operational efficiency through advanced technology [1] - Salesforce, recognized as the world's 1 AI CRM, will provide the necessary tools to facilitate better customer interactions for Haleon [1] Technology Utilization - The integration of Salesforce's AI-powered solutions is expected to drive more effective engagement strategies within Haleon's sales operations [1] - The specific tools being implemented include Salesforce Life Sciences Cloud, which is tailored for the life sciences sector, enhancing customer engagement capabilities [1]

Investment Rating - The report does not explicitly provide an investment rating for the industry or specific companies Core Insights - The global economy is expected to grow by 3% in 2025 and 2026, accelerating to 3.3% in 2027, with global inflation hovering around 2.5% [9][11] - AI is projected to drive approximately $1 trillion in spending by 2030, with over $800 billion dedicated to generative AI infrastructure [2][66] - The adoption of Agentic AI is on the rise, with an estimated spending of $155 billion by 2030, indicating a significant potential for productivity improvements [3][59] Global Economic Outlook - The global growth forecast has been upgraded by 20 basis points, largely due to China benefiting from a trade truce [9][10] - Trade policy uncertainty remains high, with geopolitical risks potentially affecting oil prices and energy importers [11][13] - The US economy is projected to grow by 1.6% in 2025-2026, reaching 1.9% in 2027, with a stable labor market [12][14] AI and Data Center Market - The global data center market is expected to reach ~$1 trillion by 2030, with AI servers representing 80-85% of the total addressable market (TAM) at ~$700 billion [2][66] - AI networking and storage are projected to account for ~$74 billion and ~$39 billion, respectively [2][66] Agentic AI Adoption - Agentic AI systems are designed to operate autonomously, with customer service, marketing, sales, and software development being the first major job functions to adopt these technologies [3][61] - Surveys indicate that 64% of organizations plan to pursue agentic AI initiatives by 2025, with significant spending potential [3][59] Precision Medicine and AI - AI is expected to accelerate the development of personalized medicine, which tailors treatments to individual patient profiles, although scalability and cost remain challenges [4][46] - Companies like Tempus AI, Guardant Health, Exact Sciences, and NeoGenomics are leading in AI precision medicine [46][48] Payments and Cross-Border Travel - A survey indicated that over 40% of respondents intend to change their cross-border travel plans, which could impact companies like Visa and Mastercard [52][53] - The travel industry is facing headwinds due to concerns about government policies and economic conditions [53][56] Semiconductor Industry - The semiconductor market is experiencing competitive dynamics among key players like Nvidia, Broadcom, and AMD, particularly in AI-related technologies [66][67] - AI data center systems are expected to grow significantly, capturing a larger share of global IT spending by 2030 [66][67]

Core Insights - The development of AI Agents is rapidly evolving, with diverse categories and application scenarios emerging despite the lack of a unified definition [6][9][42] - There are significant differences in the strategies of major companies in the US and China regarding Agent development, with North American cloud providers focusing on deployment platforms and Chinese internet companies continuing to leverage user traffic logic [2][7][42] - The high computational demand of Agent products is expected to drive advancements in the AI industry chain, suggesting a potential turning point for commercialization [8][9][42] Group 1: Agent Definition and Development - There is no clear definition of Agents, but they are categorized based on their capabilities and application scenarios, including multimodal Agents and general-purpose Agents [20][24] - Academic perspectives emphasize the need for planning capabilities in Agents, while industry views focus on the ability of Agents to independently complete tasks [10][12][18] - The evolution of Agent capabilities follows a path of "imitation learning → decoupling → generalization → emergence," enhancing their functionality across various domains [20][24] Group 2: Market Landscape and Company Strategies - North American cloud companies like Google and Microsoft are primarily focused on helping clients efficiently deploy models and Agents, while B-end companies are developing platforms for Agent creation and management [2][7] - Chinese internet giants are introducing general-purpose Agent products, while B-end enterprises are launching domain-specific Agents based on their platforms [2][7] - The commercialization of Agent products is already evident, with companies like Salesforce achieving significant revenue from their Agent offerings [2][8] Group 3: Technical Challenges and Solutions - The development of Agents faces technical challenges, including high token consumption and issues related to intent confusion and multi-Agent collaboration [2][8] - Solutions being explored include Bayesian experimental design and attention head control in academia, while industry is adopting retrieval-augmented generation (RAG) and data augmentation techniques [2][8] - Despite these challenges, Agents are demonstrating value in various applications, such as code generation and office efficiency improvements [2][8] Group 4: Investment Recommendations - The rapid progress of Agents and the upward trend in the AI industry chain suggest potential investment opportunities in software companies with data, customers, and applicable scenarios [8] - Specific recommendations include companies in ERP and government sectors, as well as those in education and healthcare that can generate new revenue streams [8] - Increased demand for model privatization is expected to benefit companies involved in integrated machines, hyper-converged infrastructure, and B-end service outsourcing [8]