# Product & Design

> What the change means for product, UX, service and interaction design, front-end development, and the accessibility that no single function owns.

Product, UX, service and interaction design, front-end development, and accessibility are changing as products become adaptive, conversational and agentic. The work spans the Product Manager, Product / UX Designer, Service Designer, Interaction Designer, UI Designer and Front-end Developer roles. Accessibility remains a responsibility without a dedicated owner.

## What changes for product & design

* **The interface learns the user.** The expectation of a zero learning curve, chat as a universal default, the user's language replacing the system's taxonomy, progressive refinement, and memory and context adaptation all land on design and front-end. See [The conversational interface](https://www.shaped-by-ai.com/en/impacts/what-is-happening/the-conversational-interface/). In concrete terms, design has to stop assuming the user learns the system before using it.
* **From showing to doing.** When the system is expected to act, design moves from laying out options to defining delegation, confirmation points and recovery paths. See [Agentic AI: from showing to doing](https://www.shaped-by-ai.com/en/impacts/what-is-happening/agentic-ai-from-showing-to-doing/). This is where the practice changes: it is no longer enough to show where to click; design must also define what the system can do on the user's behalf.
* **Serendipity becomes a design choice.** Exploratory browsing and casual discovery must be intentionally redesigned, because the system optimises for what was asked. See [Serendipity declines](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/serendipity-declines/). This forces design to decide whether and where to leave room for unrequested discoveries.
* **Intent redraws the architecture.** Someone arriving from a concrete case does not navigate the taxonomy: they need direct access by intent, not just a content hierarchy. See [Intent-based interaction](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/intent-based-interaction/). For design this means having two entry points: one for people exploring, one for people who already know what they are looking for.
* **The limit of adaptivity matters as much as its upside.** Memory and personalisation only work up to a point: beyond it, the experience turns unsettling or manipulative. See [Memory and context adaptation](https://www.shaped-by-ai.com/en/impacts/what-is-happening/the-conversational-interface/memory-and-context-adaptation/). So design has to design the upper limit too, not just the ideal personalisation.
* **The interface must not ask too soon.** Lowering the learning curve, progressive refinement and hybrid interfaces all reduce the demand for upfront training. See [The new expectation of a zero learning curve](https://www.shaped-by-ai.com/en/impacts/what-is-happening/the-conversational-interface/zero-learning-curve/) and [From form to progressive refinement](https://www.shaped-by-ai.com/en/impacts/what-is-happening/the-conversational-interface/from-form-to-progressive-refinement/). In practice, products have to stop asking for explanations before they have shown value.

> **Who holds trust calibration together.** The delegation scale, confirmation points on the irreversible, situated uncertainty signals, traceability of actions and sources, recovery paths with human escalation: five artefacts no existing role owns in full, today scattered across UX design, content design, Product Manager and legal (details in [Trust, AI literacy and calibration](https://www.shaped-by-ai.com/en/impacts/what-is-happening/cross-cutting-themes/trust-ai-literacy-calibration/)). These are service-design choices and land here, with the disclosure threshold agreed with the Product Manager. If nobody holds them together, silence decides.

## Responsibilities and actions

1. Map intents, tasks and delegation points before choosing a chat or an agent.
2. Design guided path, direct command and browse-and-discover together, without automatically replacing the UI.
3. Define accessibility criteria, states, feedback, confirmations and recovery; verify them with manual tests and real users.
4. Validate the consequences of every interaction choice with the Product Manager, content, legal and data.
5. Make functions verifiable by people and agents: semantic markup, explicit states and labels, structured data, stable selectors and tests for the main tasks.
6. Define the boundary between trusted content and untrusted instructions: sanitise third-party input and content, plan for prompt-injection signals, and keep an audit trail of delegated actions together with Operations & Data.