Sign In

AI UX Study Sharing Session 1/3

Haebom
Step 1: Intent (identifying user intent and goals)Step 2: Wayfinding (Product comprehension phase for users)Guided approachBenefitsExamplesNudge approachBenefitsImportant points when applyingExamplesSuggestionsKey features of suggestion methodsAdvantages of suggestion methodsExamplesTemplate approachFeatures of the template approachBenefits of the template approach
I once posted a blog about rethinking UX for AI (mainly LLMs), and actually ran a study group with the people who reached out. We held five sessions in total, with four participants. Next time, I think we could expand it a bit, but for now, let me share what we organized.
I'm far from perfect, but when it comes to planning, I tend to break things down into five stages. It's just my own half-baked philosophy, so please feel free to disagree or share if you have better methodologies! Like designing a typical customer journey, I try to move forward step by step considering these phases. Stage 1: Intent (understanding user intent and goals) → Stage 2: Wayfinding (user onboarding and product comprehension) → Stage 3: Using (actually using the product/service) → Stage 4: Feedback (user shares feedback) → Stage 5: Retention (retaining users and preventing churn) Now, I'll go through and organize my thoughts for each stage.

Step 1: Intent (identifying user intent and goals)

I've written about this before on my own blog. If I reinterpret it from an AI perspective, I can organize it like this. Most services today fit into the categories below.
•
Focus: Wants specific results and has a clear plan to get there (for example, editing to change the tone or format of content).
•
Navigation: Has a goal, but isn't sure how to reach it (for example, seeking an answer to a specific question but unsure how to phrase the input).
•
Synthesizing: Drawing conclusions or insights from a known data set (e.g., spotting trends or insights across different data sets).
•
Browsing: When you’re in a creative experimentation mode, trying various inputs to explore the results (like using different tokens for image generation to see what happens).

Step 2: Wayfinding (Product comprehension phase for users)

When users first encounter a product, they need to observe and understand how to use it. In reality, this process is quite heuristic and instinctive, so sometimes it's better to anticipate users’ needs or provide friendly guidance by leveraging various psychological effects. You can check the document below for psychological effects often used in UX design.
😇

Guided approach

It's not easy to craft the perfect prompt the first time you interact with AI. That’s where the 'Follow-up Prompting' approach comes in: instead of responding to the user’s initial input right away, the AI first asks for more details through an intermediate step.
1.
Simplified experience: Breaks down complicated prompts into smaller steps to reduce user burden.
2.
Build trust : Gives the impression that AI is really trying to understand the user's intent.
3.
Enhance user agency: Asking for more details encourages users to take a more active role in the process.

Benefits

•
More accessible: Even users with little AI experience can use it easily.
•
Higher accuracy: Verifying the details helps produce more accurate results.
•
Increased satisfaction: Transparent process builds trust and user satisfaction.

Examples

Advanced techniques like RAG (Retrieval-Augmented Generation) allow for an even better user experience. Thanks to these methods, users can provide direct references or parameters rather than just plain text input, saving time and effort. This is one of the most popular and easy-to-use approaches out there.
The service that makes the best use of this is Perplexity.ai . It provides a good user experience by generating follow-up questions or related questions.

Nudge approach

AI tools offer a wide range of features, but most users only take advantage of a few. 'Nudges' are a strategy for naturally introducing users to hidden AI functions and encouraging more effective use.
1.
Gradual feature introduction: Since it’s hard for users to absorb every AI tool feature at once, Nudges introduce features at appropriate moments for the context.
2.
Context-aware recommendations: By understanding the user’s current context, AI suggests features that fit their workflow.
3.
Better user experience: It shows when AI can help with repetitive or complex tasks.

Benefits

1.
Goal-oriented support: Offers features that directly help with the user’s current task.
For example: Recommending an auto-summary feature while writing a document.
2.
Support user growth: Gradually introduces complex features to help users become more proficient with AI.
For example: During data analysis, AI shows how to create a pivot table, guiding users to try it themselves.

Important points when applying

1.
Avoid overload: Instead of introducing all features at once, provide them gradually, tailored to the user’s needs and circumstances.
2.
Personalize UX: Serve different levels of nudges depending on the user's skill level.
3.
Optional activation: Let users turn nudges off or adjust them as they like.

Examples

When using Notion or GitHub , a typical example is the use of ✨ symbols to encourage the use of AI. Personally, I believe this aspect can be effectively utilized through browser extensions (an area where tools like Grammarly have excelled).

Suggestions

Suggestions are strategies where the AI suggests possible prompts or next steps to the user. This helps overcome the 'Blank Canvas Syndrome' many users feel when interacting with AI, and leads to smoother dialogue. Think of it as a way to make cold starts much easier.
😉
What's the difference from guided methods?
Guided approaches prioritize accuracy and deep understanding, while suggestion approaches focus on convenience and a continuous chat flow. Essentially, guidance involves the AI asking for extra info in response to your initial request, while suggestions involve the AI coming up with prompts or follow-up steps for you.

Key features of suggestion methods

1.
Form: Usually given as 3 to 5 clickable text options.
2.
Placement: Usually shown near the text input area.
3.
Functionality: Clicking inserts the suggested prompt automatically or produces an AI response immediately.
4.
Contextual suggestions: Suggestions reflect conversation flow and prior user actions.

Advantages of suggestion methods

1.
Easier to use:
•
It cuts down time users spend thinking about what to ask.
•
It allows users to interact with AI even without crafting complicated commands or prompts.
2.
Introducing system features:
•
It naturally introduces users to more features of the AI system.
•
It gives users a chance to discover and use features they didn’t know about before.
3.
Keep the conversation going:
•
It helps users continue the conversation when they get stuck.
•
Maintains a natural conversation flow to boost user experience.
4.
Potential for personalization:
•
By learning user preferences and behavior patterns, suggestions can get more relevant over time.

Examples

If you've used chatGPT or GPTs, you're probably already familiar with this approach. It's a useful way to spark good questions or answers.

Template approach

The template method is a powerful way to simplify complex AI systems and help onboard new users. By providing a predefined structure, users can accomplish complicated tasks with minimal input.

Features of the template approach

•
Structured template: Define the basic structure of a task in advance.
•
Flexibility: Use parameters, suggested prompts, or open-ended text in combination.
•
Automation: AI fills in the rest based on user input automatically.
•
User friendly: Lets users access advanced features without creating complex prompts.

Benefits of the template approach

•
It enables users to take advantage of advanced AI features without needing expert knowledge.
•
Even users with little experience can expect great results.
In South Korea, Rutten is systematizing this in its own way. Overseas, companies like Jasper.ai have demonstrated a good approach, and recently, they are showing various possibilities, such as conducting economic activities through prompt templates. Personally, I think of it as similar to the PPT or Notion template market.
This is getting a bit longer than I expected, so I'll continue in the next post.
Subscribe to 'haebom'
📚 Welcome to Haebom's archives.
---
I post articles related to IT 💻, economy 💰, and humanities 🎭.
If you are curious about my thoughts, perspectives or interests, please subscribe.
haebom@kakao.com
Subscribe