# From form to progressive refinement

> The user starts from an approximate idea and refines it through iterations. Tolerance for complex forms and long wizards declines.

With ChatGPT the user learns that not everything needs to be specified in advance: they start from an approximate idea and refine it through iterations. Tolerance for complex forms, long wizards, and paths that ask you to decide before exploring declines. Two patterns grow in relevance:

* progressive clarification: short questions in sequence that guide to the result, instead of extended forms;
* responsive scaffolding: contextual suggestions, inline validation, microcopy that responds in real time.

## The decision this observation informs

The decision is whether to ask for everything up front, or walk the user towards understanding what they want. It falls under the [ASSIST](https://www.shaped-by-ai.com/en/impacts/the-decisions-ai-forces/#assist) position. Without an explicit choice, everything gets asked up front, because the form is the artefact the downstream systems expect: a complete, validated record. Progressive refinement requires accepting incomplete intermediate states — that is, changing the systems, not the interface.

Progressive refinement costs longer sessions, more turns, more uncertainty about when a request is "ready", and a data model that tolerates incompleteness. The form costs the users who cannot answer the questions it asks: the ones who do not know the vocabulary, do not have the data to hand, or have not yet decided what they want. The choice belongs to the Service Designer, because it cuts across the interface and the downstream systems. This phenomenon currently involves only Interaction Designer and Service Designer, with no product or data role at all: a decision with infrastructural impact and nobody tending the infrastructure.

## Effects and role responsibilities

* **[Interaction Designer](https://www.shaped-by-ai.com/en/impacts/roles/product-design/)** — Progressive disclosure and intent-first design become central; linear forms and sequential wizards show their limits. *Progressive disclosure* (NN/g) is the pattern that defers advanced or rarely used options to a secondary level, showing only the essentials by default — the classic example is the print dialog, which opens with a handful of choices and hides the rest behind a "Show details" control. Conversational progressive refinement is a dynamic variant of the same idea: it is not the user who clicks to expand, it is the system that proposes the next question based on what it has already understood. The guiding principle stays the same: show only what is needed, when it is needed. → Useful resources: [Progressive Disclosure — NN/g](https://www.nngroup.com/articles/progressive-disclosure/); [AI: First New UI Paradigm in 60 Years — Jakob Nielsen](https://www.nngroup.com/articles/ai-paradigm/)
* **[Service Designer](https://www.shaped-by-ai.com/en/impacts/roles/product-design/)** — The journey map must foresee that the user renegotiates the need at every moment of the flow.