In a recent LinkedIn post, Hiten Shah explores the crucial role of product design in shaping the perceived intelligence and performance of AI models, particularly in the context of generative AI and AI agents. Shah emphasizes that the underlying AI model is only one part of the equation; the product surrounding it significantly influences how users experience its capabilities.
Shah highlights that the design of a product dictates how instructions are conveyed to the AI and what data it can access. This, in turn, affects the actions the AI can perform. He states:
“Product design determines which instructions reach the model and what information it can access. The available tools and permissions shape which actions it can take.”
Further elaborating on this point, Shah explains that elements like memory and verification processes are also product design decisions that impact the user’s interaction with the AI. Memory allows for decisions to be carried forward, while verification acts as a gatekeeper, determining the quality of the output that ultimately reaches the user.
The Impact of Product Design on AI Experience
Shah argues that these product-level decisions are fundamental to the user’s experience with AI. He asserts that the way an AI behaves and the quality of its output are not solely determined by the model’s inherent capabilities but are heavily mediated by the product’s architecture and user interface.
AI Agents and the Fusion of Design and Intelligence
The discussion becomes particularly relevant with the rise of AI agents. Shah points out that in the context of agents, product design becomes an integral part of the intelligence that users experience. This means that a well-designed product can make a standard AI model appear significantly more intelligent and capable.
“With agents, product design becomes part of the intelligence people experience.”
He further elaborates on the critical components that contribute to this perceived intelligence, noting:
“Memory carries decisions forward. Verification determines whether weak work reaches you.”
According to Shah, these are not minor considerations but core product decisions that shape the AI’s output before the user even interacts with it. He stresses that understanding these design elements is key to troubleshooting and improving AI performance.
Demystifying AI Behavior
To provide practical insights into these concepts, Shah announced a live session where he would dissect various AI products, including ChatGPT, Claude Code, Cowork, and Notion AI. The goal of this session, as described in his post, is to help attendees understand why these products exhibit different behaviors and how to identify the root causes when their performance falls short.
“You will see why these products behave differently and where to look when the work falls short.”
Shah’s analysis underscores the importance of a holistic view of AI development, emphasizing that user-facing product design is as critical as the underlying model development for creating effective and intelligent AI applications.
📝 About This Content
This article is based on insights shared by Hiten Shah on LinkedIn.
📅 Originally posted on August 4, 2026 | View original post on LinkedIn →