Designing a product screen is nothing like designing an AI agent that quietly decides things on its own — why not?
How does agent design differ fundamentally from interaction design practice?
This explores what changes when designers shape an AI agent instead of a traditional interface: what the actual design material is, who the design is for, and which skills carry over.
This explores what changes when designers shape an AI agent instead of a traditional interface: what the actual design material is, who the design is for, and which skills carry over. The corpus's short answer is that interaction design works on a fixed surface the user learns, while agent design works on a shifting layer of context, instructions and coordination that the user never fully sees. Dubberly goes further and argues that agent design will absorb interaction design rather than sit beside it. He treats LLMs as a new design material, comparable to earlier technology shifts that reshaped whole design fields Is design for agents becoming the next major design paradigm?.
The clearest break is about context. A traditional UI stays put: buttons stay where they are, so users build a mental model and rely on it. For an AI system, the 'interface' is a mix of prompt, conversation history, retrieved data and hidden state that changes with every turn. Users can't learn it the way they learn a menu. That's why one note argues the core discipline moves from interface design to context engineering How does AI context differ from conventional software context?. Look at the skills Salesforce's Kat Holmes names for AI-era teams: experience architecture, agent collaboration, conversation design, and context maps and ontologies. Only one of them, conversation design, looks much like classic interaction design What design skills do teams need for the AI era?.
The second shift is that much of agent design never appears on a screen. Dubberly describes it as stacking capabilities (chatbot → system prompt → retrieval) and layering oversight (monitoring agents → supervisors → orchestrators) How should agent capabilities be organized into functional layers?. Engineering research comes to a similar view from another direction. Reliable agents come less from bigger models and more from moving memory, skills and protocols into a surrounding 'harness' Where does agent reliability actually come from?. In multi-agent systems, structured documents passed between agents coordinate work better than open-ended chat Does structured artifact sharing outperform conversational coordination?. Fin's practitioners make the same point at small scale. Facts belong in a retrievable knowledge base and only tone and behavior rules belong in the instructions, so deciding where information lives becomes a design decision Should factual content and behavior rules live separately?.
The third shift is the most surprising: the 'user' may not be a person. Interfaces are now being designed for agents themselves. Adding agent-specific heuristics about perception, grounding and control helped computer-use agents finish more tasks, and human usability didn't get worse Can interface design help agents complete computer tasks better?. Delegation can also run in reverse, with agentic systems assigning tasks and responsibilities to humans Can AI systems delegate tasks back to humans?. Beyond the single product, Chan et al. argue that agents need shared infrastructure, like internet protocols, to attribute actions, shape interactions between agents, and fix harm. That is design at the level of the ecosystem rather than the screen What infrastructure must exist beyond training agents?.
The old craft still matters, though. Conversation design remains a lever you can measure: one warmer opening message raised customer satisfaction from 72.8% to 78.4% in an A/B test Does conversation design ownership measurably improve AI agent performance?. Classic HCI principles also still separate good agents from bad ones. In a review of nine deployed agents, general-purpose tools scored well, while companion and advisory agents most often broke principles like preventing dependency and making the agent's intent visible Do productivity agents actually protect users better than emotional ones?. So the difference isn't that interaction design stops mattering. It becomes one layer inside a bigger job: designing what the agent knows, how it coordinates, and who it answers to.
Sources 12 notes
Dubberly argues design is entering a new paradigm driven by LLMs as a design material, where agent design will subsume interaction design and become core to service design. He grounds this in design history's pattern of technological shifts reshaping design disciplines.
AI interactions operate on a substrate of constantly shifting context—prompt, history, retrieved data, hidden state—that users cannot internalize like traditional UIs. This structural mutability demands a new design discipline centered on context engineering rather than interface design.
Salesforce's Chief Experience Officer identifies experience architecture, agent collaboration, conversation design, and context maps and ontologies as the design competencies teams must develop. Each skill addresses a distinct gap in current AI system design.
Dubberly proposes a two-axis stack model: a capability axis (chatbot → system prompt → RAG) adding inputs that extend what LLMs can act on, and a scale axis (monitoring agents → supervisors → orchestrator) aggregating oversight. He cautions that ubiquitous agent hierarchies raise design risks alongside capability gains.
Research shows reliable LLM agents externalize three cognitive burdens—memory (state persistence), skills (procedural components), and protocols (structured interaction)—into a harness layer rather than relying on model scale alone. The harness unifies these externalities and eliminates the need for the model to solve the same problems repeatedly.
Show all 12 sources
MetaGPT demonstrates that agents producing standardized engineering documents achieve superior coordination compared to conversational exchange. Active information pulling from shared environments eliminates noise and mirrors efficient human workplace infrastructure.
Fin's Blueprint Team recommends routing factual content to retrievable knowledge bases while keeping only tone and behavioral directives in agent instructions. This split aligns with instruction-tuning research showing models follow what-to-do rules better than what-to-ignore constraints.
UI-Verse benchmark testing found that augmented heuristics targeting agent perception, grounding, and control consistently improved agent task completion and efficiency. Human studies confirmed these designs preserved usability for human users with no observable regressions.
Holldack, Banh and Strobel argue that agentic information systems capable of independent decision-making can transfer rights and responsibilities to humans, reversing the traditional one-directional delegation model. This requires moving beyond agency-primacy frameworks to delegation theory that accounts for bidirectional task assignment.
Chan et al. argue that external technical systems modeled on internet protocols are essential beyond behavior training. These systems must handle three core functions: attributing actions to agents, shaping their interactions, and detecting and remedying harm.
Fin's Blueprint Team found that intentional conversation design—controlling tone, detail level, and handoff structure—measurably improves outcomes. A single warm opening message raised CSAT from 72.8% to 78.4% in A/B testing.
A heuristic evaluation of nine deployed agents found that general-purpose productivity systems (ChatGPT, Claude) scored substantially higher on 14 HCI principles than emotionally sensitive or high-stakes agents (Character.ai, Consensus, Pine AI). Risk-critical principles like dependency prevention and intent visibility were most violated in companion and advisory contexts.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- A Framework of User Experience Principles for Human-AI Agent Interaction in the Workplace
- Design Principles for Human-Agent Interaction
- Assistant or Actor? Student Trust, Control, and Delegation Regret When Using a General-Purpose AI Agent
- 2025 in Review: Design Gives AI Meaning
- A Framework for Studying AI Agent Behavior: Evidence from Consumer Choice Experiments
- Intelligent AI Delegation
- Embodied LLM Agents Learn to Cooperate in Organized Teams
- Choose Your Agent: Tradeoffs in Adopting AI Advisors, Coaches, and Delegates in Multi-Party Negotiation