2025.06. 24 丰农控股集团总农艺师王兴林表示，近五年时间，团队在AI方面的实践包括了图像识别、病虫害管 理、生产力评估、农场综合管理、精准施肥用药等方面。我国农业文明虽然悠久，但工业化冲击对农 业发展带来的提升空间，目前来看仍然很大。也因此，丰农的定位仍以农业服务为核心，通过AI工 具属性来服务农业发展。 AI农业智能化提升慢的原因可以归结为三大方面：数据基础薄弱、高成本与低ROI预期、场景复杂性 与信任壁垒。由于土壤、气候、作物品种多样等原因，农业数据分散，农业相关数据集不足以训练可 靠AI模型，且缺乏实时数据收集能；智能农机、无人机等农业AI设备成本高昂，农业行业投资回报 周期长；从业者更信任经验，将AI视为辅助工具。 本文字数：1342，阅读时长大约3分钟 作者 | 第一财经 吕倩 6月24日，中金公司研究部发布的研报显示，AI渗透需长期演进，尤其从技术萌芽到企业大规模部署 阶段，会经历资源分配、用户习惯转变、技术与场景打磨、产品迭代等节点，可能存在短期阵痛。今 日收盘，AI应用(8841683)指数收涨2.55%。一方面，大模型落地行业应用进展迅速，但同时，部分 领域AI渗透率仍增长缓慢。A ...

Core Insights - The report from CICC indicates that AI penetration in agriculture requires long-term evolution, facing challenges such as resource allocation, user habit changes, and technology refinement [1] - AI applications in agriculture are currently in the experimental demonstration phase, with slow penetration rates in certain areas despite rapid advancements in large model applications [1] - The complexity of agricultural environments and the need for precise recommendations pose significant challenges for AI models, which must adapt to various conditions [7][8] Industry Overview - The agricultural sector has seen the emergence of multiple large model products, with competition centered around data accuracy and relevance to customer needs [4] - The investment landscape in smart agriculture has been active from 2015 to 2022, yet only 20-30 companies receive funding each year, indicating a selective investment environment [8] - The Ministry of Agriculture and Rural Affairs has set a goal for agricultural production informationization to reach over 32% by the end of 2028, highlighting the importance of national policies in fostering AI agricultural development [8] Challenges in AI Agriculture - Key challenges hindering AI adoption in agriculture include weak data foundations, high costs, low ROI expectations, and trust barriers among practitioners [1] - The diversity of agricultural data due to varying soil, climate, and crop types complicates the training of reliable AI models [1] - The complexity of agricultural environments limits the generalizability of AI solutions, making it difficult to apply models developed in one context to another [7][8]

Group 1 - The core viewpoint of the article highlights the successful trial of high-quality crop varieties developed by the "Shennong·Guxin" smart breeding model, marking a new phase in the industrial application of smart breeding technology [1] - The "Shennong·Guxin" model integrates genotype data and agricultural knowledge graphs, optimizing the breeding process and reducing the breeding cycle by over 30%, from the conventional 5-8 years to 2-3 years [1] - The smart farm in Yanqi Town features 18 acres of experimental fields designed for immersive agricultural experiences, while the Danhui Agricultural base showcases improved yield and reduced pesticide use through smart breeding [1] Group 2 - Smart management systems based on the "Shennong model" have been implemented in multiple districts of Beijing, with the use of smart inspection robots and drones in Changping District leading to a 30% reduction in labor costs and a 10-15% decrease in fertilizer and pesticide usage [3] - The "Shennong Digital Human" technology facilitates human-machine dialogue for agricultural knowledge and production guidance, applied in farm research centers and smart greenhouses in Haidian District [3]