Understanding
What problem counts?
- Use and context
- Technical rules
- Open assumptions
Product work reduces uncertainty step by step. A clear problem, tangible designs and joint decisions give the team orientation.
Discovering LeadershipFour typical starting situations and the appropriate next steps.
Where does the user lose the expected value?
What change is most important now?
What is missing for the next independent decision?
Which assumption must work first?
The steps provide guidance. Feedback can return to an earlier step at any time.
What problem counts?
What does the team need to know?
How does it become reliable?
What has changed?
Prompt-based design work combines requirements, gross layout and implementation. Product and design set the direction; agents condense the context and formulate the next working and design prompts from it.
At the beginning are the business problem, the people affected and the desired result. This results in a structured work order with framework conditions and verifiable acceptance criteria.
The agent structures the business goal in terms of user problem, process and acceptance criteria. Open questions become visible before the draft solution.
Which problem does the product solve – and how can a good solution be identified?
Preparatory work.
From the briefing, agents develop first layout variants and a clickable prototype. The Groblayout shows the idea in context: entry, core action and next step. So the basic direction can be tried early.
The agent develops two simple layout variants for the core process. The most important interaction is clickable; the assumptions for each variant are named.
Do people understand the process even before detailing it?
Based on the chosen direction, prompts sharpen the information hierarchy, navigation and interactions. Wireframes also make the less visible states tangible: empty, loaded, faulty or without authorization.
The agent works out the selected sequence as wireframes. Load, blank and error states, mobile views and keyboard operation are considered.
What happens before, during and after the actual action?
In the fine layout, wireframes combine with typography, distances, colors and existing components. Prompts reference the design system and its variants. Design examines visual hierarchy, comprehensibility and accessibility related.
The agent transfers the wireframes into the existing design system. It uses semantic tokens, referenced components and their application rules. Responsive variants, focus and error feedback are considered.
Does the system carry the solution on small and large screens?
<Button /><Input />UI + Behaviour + DataReviewTests + Browser ComparisonSame system. Less translation.
The coding agent receives the design context, the associated code components and the acceptance criteria. A connection such as Figma MCP and Code Connect can support this transfer. Behaviour, data and test orders are just as much a part of the specification as the visible layout.
The agent implements the flow with the existing code components. Conditions, data contracts and acceptance criteria form the basis. Function and presentation are checked in the browser.
Does the current product correspond to the intended use?
The context is preserved. Requirements, decisions and feedback flow into the next prompts. The team clarifies open questions about the direction together on the draft. For design experiments and tests with people, there is conscious space.
When design system, codebase and quality assurance work together, product can orchestrate the way to a workable proposal. This target image describes the process in an appropriately set up product organization.
Define goal, user problem, scope and acceptance criteria. A briefing agent structures the tasks and prompts from this; product clarifies open decisions.
A versioned briefingGroblayout and prototype development, complete wireframes and transfer them to the design system. Behavior and conditions remain part of the same specification.
UI, Flow and Component ReferenceUse existing components and interfaces. Implement the change in your own branch and supplement suitable tests to the acceptance criteria.
Code change with testsCheck build, types and tests. Design checks control system fidelity; browser checks complement critical user paths, presentation and accessibility. Findings go back to the agents.
Test report instead of ready reportProvide the tested stand in a separate environment. Preview link, screenshots, change overview and open points make the result testable together.
A running staging versionDesign Head and Engineering Head will jointly review the current results. Product takes the requirements technically. Feedback becomes concrete follow-up orders.
Release or next iterationFeedback becomes the next order. Findings from checks and reviews go back with context to design or coding agents.
For clearly defined changes, the formal design and engineering review can be bundled at the end. Both heads test UX, system reliability and technical quality; product is responsible for professional acceptance. Context, comprehensibility and impact require human judgment. Greater uncertainty needs earlier voting. The release in production remains a decision of its own.
Semantic tokens, code components and versioned rules form the common context. Open formats keep it independent of the individual tool.
Goals, acceptance criteria, data contracts and decisions are versioned. Open technical questions clarified product.
Appropriate test data, automated checks and reproducible preview builds make changes verifiable before release.
A nominated owner is responsible for quality, operation and further development. DesignOps maintains rules and checks; contributions are reviewed and versioned.
Collaboration starts where product, UX and technology come together: in building new products, in growing platforms and with teams that want to further develop their collaboration.
The team ranks the challenge, the people involved and the most important open question.
A common focusAn existing process, existing data or a first prototype make the question tangible.
A verifiable assumptionThe next meaningful step, the responsibility and the feedback are agreed together.
A clear next step