Useful AI Features: Value, Prompts, Visibility

Key Takeaways

1The value of an AI feature must be demonstrated beyond the simple label "AI"

2Prompt writing assistance enhances usability and exploration of tools

3Outputs must be readable, specific, and easily discoverable

4An agent tested in 2025 was found to be slow and fallible

Why it matters : Designing useful AI features requires addressing real needs, supporting the user, and making capabilities accessible and understandable.

A corpus of recommendations details how to design AI features that truly serve user needs. It emphasizes explicit value, assistance in prompt writing, output readability, and discoverability, while reporting on limitations recently observed in the use of agents.

Observed Limitations: Slowness and Errors in an Agent in 2025

In 2025, a ChatGPT conversational agent successfully booked a table at a restaurant, but the process was slow and prone to errors. Some recommendations made at one point remain valid, while others have likely evolved with the rise of AI capabilities and their adoption in UX. Furthermore, the case of Rufus at Amazon shows that even AI features deemed useful can have little impact if they go unnoticed by users.

Prioritizing Needs: Technology Alone is Not Enough

As early as 2021, Jakob Nielsen warned against over-investing in technology without a clear understanding of user needs. The recommendations stress the importance of focusing on real needs rather than novelty. The label "AI-powered" does not constitute a value proposition in itself: the value provided must be made explicit. Features must address real problems, and the adoption of a chat interface is only relevant if it meets a user need. Narrowly focused AI features are easier to understand and adopt.

Facilitating Prompt Creation Improves Tool Adoption

The effectiveness of features based on language models largely depends on users' ability to formulate relevant prompts. Even though recent models can produce useful results with minimally detailed queries, they cannot guess implicit intentions, and users do not always know what they are looking for or what the AI can accomplish. Therefore, it is essential to support prompt writing. Interfaces that combine prompt input with graphical elements enhance usability, particularly for image generation. Well-designed prompt suggestions foster learning, stimulate creativity, help explore capabilities, and set realistic expectations. In chatbots, prompt controls serve four main functions: increasing discoverability, educating and inspiring, establishing constraints, and facilitating follow-ups.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Output Rules and Well-Defined "Superpowers"

AI is attributed four main strengths: content generation, synthesis, basic data analysis, and the ability to adopt different perspectives. Users value AI-generated summaries of opinions when they present a sufficient level of detail and clearly indicate their sources. For web writing with GenAI, results must remain concise, easy to skim, structured according to the inverted pyramid, and written in clear language. AI can potentially better understand user intentions and increase the productivity of UX professionals, provided these requirements are met.

Resources and Feedback on Design and Ethics

Several resources address strategy, prompt writing assistance, product recommendations, and the state of AI product design. Henry Modisett, head of design at Perplexity, shares tips for designing AI features in a podcast episode (number 36). Paige Lord, Senior Product Marketing Manager at GitHub, explains how UX professionals can contribute to responsible AI products and services in episode 43.