Scenario-Specific AI

Scenario-specific AI is [[ZhangQi|张奇]]’s product rule in Vol.114 AI的2025和DeepSeek们的未来 | 对谈复旦张奇教授: the core unit of AI value is a concrete scene, not a broad vertical industry label. In the episode, Cursor works because it is optimized around code, project context, and developer expectations; Perplexity works because it is optimized around search and answer gathering rather than generic chat.

The concept overlaps with Vertical Workflow AI, but it is narrower and more diagnostic. A “vertical” can still be too broad if it does not specify the job, user type, inputs, outputs, review standard, and downstream action. Scenario-specific AI asks whether the model, data, interface, and evaluation loop are built around the real moment where the user needs help.

Key Claims

• A strong general model is often not enough; the product has to know what good output means in the target scene.
• User segmentation matters inside one domain: beginner programmers and professional programmers may need different AI coding products.
• Prompt quality can matter, but the bigger difference often comes from training, tool access, context, and workflow design.
• The most defensible AI products may be those that reduce a specific user’s work steps while preserving review and correction.
• Contact Center AI, AI search, and coding assistants are promising because their scenarios have repeated tasks, observable outcomes, and clearer acceptance criteria.

Connections

• Vertical Workflow AI and AI Application Layer Moat — broader product-defensibility frame.
• Model Workflow Fit and Model Routing Cost Control — model selection by task and workflow.
• Cursor, Perplexity, ChatGPT, and DeepSeek — product and model examples from the source.
• AI Programming Engine Shift, AI Coding Verification, and Human Judgment Under AI — coding and review implications.
• Contact Center AI, Customer Support Automation, and Product Led Willingness To Pay — deployment and business-value cases.