If an AI quietly sets its own goals as it talks with you, does a human need to approve every single move?
Do just-in-time objectives require active designer judgment at every interaction?
This explores whether a 'just-in-time objectives' setup, where an AI works out what it is aiming for during each interaction, needs a human designer making calls every time, or whether that judgment can be built into the system itself.
This explores whether a 'just-in-time objectives' setup, where the AI's goals are generated and checked live during each interaction, needs a designer weighing in every time. The corpus suggests the answer is no, but with a catch: the judgment doesn't go away, it changes form. In the argument by Lam and colleagues, the point of the architecture is that it makes the generate-then-evaluate steps visible at every interaction instead of hiding them inside the model Does visible design structure keep HCI expertise necessary as AI advances?. So the designer isn't approving each turn. They are shaping the structure that does the approving: what gets generated, what counts as a good result, and what the user gets to see. Design judgment becomes something the system runs, not something a person does live.
Why not just let the model handle it? Other notes show what happens when design intent is left to the AI. One benchmark found that generative UI tools quietly dropped about a quarter of the design reasons they were given, and about a third of the functional requirements Do generative UI tools actually implement their stated design rationales?. Fin's team found that something as small as one warm opening message measurably raised customer satisfaction, which is the kind of choice a model won't reliably make on its own Does conversation design ownership measurably improve AI agent performance?. Together these suggest that judgment must enter the system somewhere. Just-in-time objectives put it into the evaluation step instead of trusting the model's defaults.
The less obvious point is that every interaction needs fresh judgment because what the AI is working with keeps changing. Its prompt, history and retrieved data shift from one moment to the next, unlike the fixed screens of traditional software How does AI context differ from conventional software context?. What an AI 'remembers' is itself a product decision about which stored memories get pulled in How do AI products decide what to remember between conversations?. A design fixed once at launch can't keep up with that. That is why the corpus keeps pointing toward ongoing, built-in evaluation and away from one-time specs. Older frameworks already had a 'during interaction' category How should AI systems be designed across different interaction phases?. Just-in-time objectives effectively turn that category into a running process.
There's also a second answer to 'whose judgment?': sometimes it moves to the user. Cocoa lets people assign plan steps to themselves or the agent and revise the plan mid-task. Users got more control without it costing them extra effort Can users steer AI agents better by revising plans mid-execution?. Tools like Canvil go the other way, giving designers a low-barrier way to shape model behavior directly through prompts Can designers shape LLM behavior without deep technical knowledge?. The emerging picture is that designers write the rules for judging, while users and systems apply them moment to moment. That changes what designers need to be good at. Kat Holmes names context maps and agent collaboration as core skills for the AI era What design skills do teams need for the AI era?, and those are skills for designing a structure, not for watching individual turns.
A caveat: the corpus has one direct source on just-in-time objectives. It has no evidence on how much ongoing designer oversight these systems need in real deployments. The claim that the judgment can be built in is an architectural argument, not a measured result.
Sources 9 notes
Lam and colleagues argue that just-in-time objectives architecture makes design decisions between AI and interfaces visible, ensuring HCI expertise remains critical. By surfacing generation-and-evaluation steps at every interaction rather than hiding them, the architecture structures ongoing design judgment into the system itself.
A benchmark of 24 tasks across five tools found roughly 25% of design rationales go unimplemented, rising to 34% for functional requirements. Tools recognized only half the UX principles embedded in prompts.
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.
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.
Nielsen Norman Group's glossary shows that context is dynamically assembled from multiple sources—including stored memory—for each request. Because products control which memories get surfaced, claims about what an AI remembers describe the product, not the underlying model.
Show all 9 sources
Microsoft researchers synthesized over 150 design recommendations into 18 actionable guidelines grouped by when they apply: initially, during interaction, when wrong, and over time. Validation with 49 practitioners across 20 products confirmed their applicability.
Cocoa's system, which lets users assign plan steps to themselves or an agent and revise plans mid-execution, afforded greater steerability than chat-baseline interaction in a lab study (n=16) while maintaining ease of use, with field users valuing explicit work delegation.
Canvil demonstrates that designers can effectively shape LLM behavior via a low-barrier Figma widget for prompt authoring and testing, bringing user-centered judgment directly into model adaptation without requiring engineering expertise.
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.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- What does Generative UI mean for HCI Practice?
- Guidelines for Human-AI Interaction
- A Framework of User Experience Principles for Human-AI Agent Interaction in the Workplace
- Design Theater: Evaluating the Gap Between User-Facing Design Reasoning and Implementation in Generative UI Tools
- Just-In-Time Objectives: A General Approach for Specialized AI Interactions
- The Designer's AI Glossary: Familiar Words, New Meanings
- 2025 in Review: Design Gives AI Meaning
- Prototyping Multimodal GenAI Real-Time Agents with Counterfactual Replays and Hybrid Wizard-of-Oz