AI Feature Discoverability: Best Practices & Common Pitfalls

Over the past three years, since OpenAI launched ChatGPT, it's become nearly impossible to talk about technology or product development without the conversation eventually turning to GenAI. OpenAI has also made it easy for everyone to launch ideas based on AI technology. Investors are increasingly funding AI-based startups, creating a fertile environment for growth.

But what about large companies that already have established products and a loyal customer base? No one wants to repeat the story of Nokia or Kodak, companies that failed to recognize a new technological reality and lost their market position as a result.

As a user, I've noticed that almost every product I use has implemented some form of AI. As a product manager, I became curious about the discoverability of these AI features: how easily can users find and use them? What new capabilities are available to me as a customer? This curiosity led me to conduct my own product research.

For my research, I chose three products: Figma, Amazon and Duolingo. All three heavily promote their AI features and build strong awareness, making them interesting case studies for analysis.

Let's explore what makes AI features discoverable and how to avoid common pitfalls, even those that large companies face.

What is AI Feature Discoverability?

Let's start with the basics. Discoverability refers to how easily users can find features, information, or functionalities within a product. In UI/UX design, it's about making key tools and options visible or easily accessible when needed.

💡 "It's not enough to simply create a valuable feature; you also need to create awareness and drive adoption. Even the most useful AI features won't be used if people don't notice them."

According to Figma's 2025 AI report, 34% of users say they've shipped applications with generative AI, up from 22% last year. Additionally, 56% report their companies are integrating AI into existing products, and 43% are creating entirely new products with AI capabilities. These insights likely reflect broader market trends: more companies are adding AI, and more users are interacting with AI-powered tools.

Today, it's trendy to announce that your product "has AI." This sets high expectations. I'm always curious to test what's new. However, in many cases, I find that these features don't meet expectations. I even ran a small survey among people around me. The sentiment was similar: AI features often don't perform perfectly yet, but people believe they will improve with time.

Case Study #1: Figma

Figma has significantly expanded its AI capabilities, integrating tools that streamline workflows, enhance creativity, and automate repetitive tasks.

🛠️ AI Features in Figma Design:

• First Draft: Generates editable design layouts from text prompts.

• Visual & Asset Search: Allows users to search for assets using images or descriptive text.

• Replace Content: Fills in realistic content in place of placeholders.

• Add Interactions: Suggests and adds interactive elements to prototypes.

• Rename Layers: Uses AI to rename layers based on content.

• Text Editing Tools: Includes rewriting, translating, tone adjustment, and summarizing.

• Image Editing Tools: Offers background removal and resolution enhancement.

✍️ AI Features in FigJam:

• Generate Boards and Diagrams: Creates structured diagrams from text input.

• Sort and Summarize Stickies: Organizes sticky notes automatically.

Despite this, I couldn't find any AI features available in Figma Design: not in my corporate account, nor in my personal one. I even checked the pricing plans and documentation but found no clear explanation. This was disappointing. In contrast, FigJam did have discoverable AI features. The AI button was clearly visible, and I quickly understood that it offered some AI-driven experience.

Case Study #2: Amazon

Amazon has long used AI behind the scenes: think recommendation engines, dynamic pricing, or supply chain optimization, but more recently, it's begun promoting AI features more openly to users.

🛒 AI Features on Amazon:

• Product Recommendations: Powered by collaborative filtering and behavioral data, the "Recommended for You" section adapts based on your browsing and purchase history.

• Review Summaries: Amazon now uses generative AI to summarize product reviews into short bullet points. This helps users quickly understand product pros and cons.

• Alexa Shopping Assistant: Integrated voice assistant that suggests products, reorders items, and answers user questions based on past behavior and context.

• "Customers Ask" Feature: Auto-generated answers to common questions based on existing reviews and Q&A content.

From a user discoverability perspective, Amazon is strong in integration, but weak in transparency. Many users benefit from AI features without realizing they are AI-powered. For example, the review summary is helpful, but there is no clear label that it's AI-generated, nor an option to dive deeper into how it was produced.

This creates a situation where the value is delivered, but user awareness and trust may lag behind. Transparency is key if you want users to develop confidence in the feature and rely on it actively.

Case Study #3: Duolingo

Duolingo is a great example of transparent, engaging AI implementation. The company has embraced generative AI both in the backend and as a direct feature for learners.

🧠 AI Features in Duolingo:

• Smart Tips: Contextual grammar tips and nudges personalized to each learner's performance.

• Dynamic Review: AI suggests review sessions based on previous mistakes and knowledge gaps.

• Duolingo Max: A paid tier that gives access to two AI-powered features:

  • Explain My Answer: Offers a conversational explanation of why an answer is right or wrong.

  • Roleplay: Uses GPT-powered conversations with fictional characters to practice speaking in real scenarios.

From a discoverability standpoint, Duolingo excels. The AI features are clearly labeled, playfully designed, and placed right in the learning flow. When users make a mistake, they see an "Explain" button. When roleplay becomes available, it's introduced through a fun character like a waiter or a travel agent.

The company also uses animations and humor to lower the barrier of trying something new, which is a best practice in UX when introducing AI-driven functionality.

My Recommendations

Based on these observations, here are some best practices and pitfalls to consider when designing AI feature discoverability:

✅ Best Practices

1. Make It Visible: Don't hide your AI features in submenus or ambiguous labels. Use icons, buttons, or onboarding flows to highlight them.

2. Set Realistic Expectations: Clearly explain what the feature can do and what it can't. Avoid the "magic" trap.

3. Provide Contextual Triggers: Integrate AI features where users need them most (e.g., after making a mistake, while searching, or when idle).

4. Allow Exploration: Let users test, undo, and play with AI features safely. People learn by doing.

5. Be Transparent: If something is generated by AI, say so. And ideally, explain how it works in simple terms to build trust.

6. Track Usage and Feedback: Monitor engagement and listen to users to improve discoverability over time.

❌ Things to avoid

• Overhyping Without Delivery: Announcing AI features that underdeliver damages trust.

• Confusing or Hidden Entry Points: If users can't find the feature, it doesn't exist in practice.

• Lack of Contextual Relevance: Pushing AI tools when users aren't ready for them leads to rejection.

• Poor Explanation of Value: If the benefit isn't obvious or immediate, users will ignore the feature.

The success of an AI feature doesn't just depend on how well it performs - it depends on how well users can discover, understand, and trust it. As product builders, we need to go beyond shipping functionality and invest in how we present AI within our user experiences. Discoverability is not a "nice-to-have", it's a core part of the product's value delivery.