Intuitive AI (Part 2): UX Guidelines for Gen AI Products, Putting Theory into Practice
Adi Mazor Kario
Tech Product, Design & AI Expert ?? l Empowering Tech Companies to Lead with Unbeatable Products l Google Gen AI Expert l Keynote Speaker l Bestseller Author
### Translation to English:
Designing an exceptional user experience (UX) for generative AI (GenAI) products requires a deep understanding of AI technology alongside profound insights into the users and customers. The main challenge is to find the perfect balance between leveraging the potential of AI and genuinely addressing user needs. Let's dive into the essential elements that drive effective UX design in AI-based products and understand how they impact the overall user experience.
Following the first post about Intuitive AI, let's dive into some practical guidelines with real-world examples. These principles will help you create AI experiences that are intuitive, efficient, and truly user-centric.
We will focus on the following guidelines:?
Orientation and Navigation
Guideline: Ensure users always know where they are in the flow and what their next options are. This is true in visual interfaces and with conversations- users want to know where the interaction is going to and what are the options on the way.
Example: Perplexity AI enhances the user experience compared to the original ChatGPT, even though both utilize OpenAI's language models. It offers clear “Sources” citations for responses, allowing users to trace information back to its sources, which builds trust and facilitates deeper research. The "Related" section suggests follow-up topics based on the current conversation, encouraging users to explore subjects more comprehensively. Additionally, the "Rewrite" feature allows for easy refinement of answers, enabling users to obtain more precise or detailed information effortlessly. These orientation and navigation features create a more intuitive and productive environment for information seeking and exploration.
Ensure users always know where they are in the flow and what their next options are.
Use Gen AI Capabilities wisely
Guideline: Only require users to write or ask when necessary. If you can predict potential next actions or conversation topics, offer them directly.
Example: Trender.ai's recent addition of an AI-powered "style assessment" feature for prospects' LinkedIn profiles raises questions about the practical value of such capabilities in sales enablement. While this feature showcases the ease of implementing AI-driven analysis, its relevance to actual sales processes is debatable. Instead of focusing on writing style, AI tools in sales should prioritize more pertinent questions that directly impact the sales approach. For instance, identifying the most important topics relevant to the prospect in relation to the offer, or determining the optimal timing and method for making an introduction that will elicit a response. These insights would provide salespeople with actionable intelligence, enabling them to tailor their approach more effectively and increase their chances of successful engagement. The UX of AI tools (in this case, sales tool) should be guided by practical needs rather than the mere availability of technology.
The UX of AI tools should be guided by practical needs rather than the mere availability of technology.
Provide the Fastest Path to Solutions
Guideline: Writing isn't always the quickest route. Offer alternative, faster methods when possible. Make the users lives as best and easy as possible.
Example: Thomson Reuters' Westlaw Precision AI, despite involving 2000 legal experts, 4500 technology experts, 125 AI engineers, and data scientists, and an investment exceeding $100 million, exemplifies poor UX design for AI applications. The tool, designed to assist legal professionals by providing relevant case references, fails to offer a coherent workflow. Ideally, users should be able to view his cases, select the relevant one, ask questions, receive easy-to-understand summaries and reference cases , and seamlessly add references to their case. However, the current system treats each query in isolation, lacks contextual continuity, and requires users to manually type requests to save references, instead of providing intuitive buttons. This disjointed user experience highlights a significant oversight in integrating effective UX principles, despite the substantial resources and expertise involved.
Writing isn't always the quickest route. Offer alternative, faster methods when possible. Make the users lives as best and easy as possible.
Simplicity is Key
Guideline: Offer the most straightforward path for users to achieve their desired result, view, or action.
Example: Intuit QuickBooks quickly integrated Gen AI chat into their dashboard, but discovered low user engagement due to uncertainty about how to interact with the AI. To address this, they added conversation chips or pre-set questions to initiate dialogue. While this approach is similar to traditional button links for reports and features, it serves as a starting point for the new AI feature. However, it doesn't significantly differentiate from typical chatbot experiences. There's potential for further optimization to fully leverage the capabilities of Gen AI and improve user interaction.
Offer the most straightforward path for users to achieve their desired result, view, or action.
Context-Sensitive Solutions
Guideline: Understand the user's goals at each step and tailor the experience accordingly. Don't rely solely on AI/LLM capabilities.
Example: Salesforce has successfully integrated AI features across their product suite, with their AI assistant Einstein showcasing exemplary implementation in the context of client interactions. Unlike Trender AI, Salesforce's UX and product teams have excelled in their approach. They provide sales representatives with a comprehensive set of relevant data to enhance sales effectiveness. This includes a client's purchase history, spanning from wedding registry to home purchase and baby registry, as well as insights into their style preferences, recent views and purchases, and likelihood to buy. This wealth of information is precisely what a sales representative needs to engage effectively with a customer and drive sales. Furthermore, the interface offers easy access to relevant actions such as updating information or initiating a call, streamlining the sales process. This thoughtful integration of AI and user experience demonstrates Salesforce's commitment to creating tools that genuinely support and enhance the sales process.
Understand the user's goals at each step and tailor the experience accordingly.
Familiar But Better
Guideline: Base the user experience on concepts and flows users are already familiar with, but enhance them with AI capabilities.
Example: The best example of using a familiar flow and environment in the context of Ai is the code creation capabilities in GitHub copilot. This is within the original IDE environment, the engineer is doing the same tasks - going into Github repository and taking the right components to integrate into the code- all of this is the same familiar usage, just faster and better. The change in productivity of engineers is very substantial and both GitHub and outside researchers have observed positive impact in controlled experiments and field studies where Copilot has conferred: 55% faster task completion using predictive text. Quality improvements across 8 dimensions (e.g. readability, error-free, maintainability) 50% faster time-to-merge.
Base the user experience on concepts and flows users are already familiar with, but enhance them with AI capabilities.
Conclusion
As we wrap up this exploration of UX design for Gen AI products, it's clear that we're standing at the threshold of a revolutionary era in human-computer interaction. By implementing these guidelines, you can create AI-powered experiences that not only harness the incredible potential of artificial intelligence but also feel intuitive and natural to users. Remember, our ultimate goal is to amplify human capabilities, and these UX principles are your compass in navigating this delicate balance.
But this is just the beginning. As we venture into the realm of multi-modal AI - systems that can read, write, see, create visuals, hear, and generate voice and music - we're entering a new frontier of holistic UX design. The future promises AI interfaces that engage all our senses, offering unprecedented levels of interaction and creativity.
As we wrap up this exploration of UX design for Gen AI products, it's clear that we're standing at the threshold of a revolutionary era in human-computer interaction. By implementing these guidelines, you can create AI-powered experiences that not only harness the incredible potential of artificial intelligence but also feel intuitive and natural to users.
What Do you think?
I'm incredibly excited about the possibilities this future holds. What about you? What innovative AI interfaces have you encountered that hint at this multi-modal future? How do you envision UX evolving as AI becomes more versatile and perceptive? Share your thoughts, experiences, and predictions in the comments below. Let's continue this fascinating conversation and shape the future of AI-driven user experiences together!