UX of AI - Reality Distortion Unraveled

If you've seen discussions suggesting AI products can do away with a traditional UI, replaced by a simple search box, you're not alone. But is this trend realistic? Let me share why UI is still essential for AI-driven products. Does this example image of a SaaS offering that clearly says "No more menus" looks familiar?

But is this a realistic vision? Having built two companies around search for enterprise data, I’d argue it’s not. Here’s why the regular nuisances of UI are still essential, even with an advanced search function.

1. The Home Page is an Absolute Must

A home page isn’t just convenient—it’s essential. A business user will not start a typical day thinking about what to ask in the AI platform. Users often log in with specific goals in mind, like checking daily metrics or performing routine tasks. A home page serves as a familiar starting point, displaying key metrics and guiding users without requiring them to type every inquiry into a search bar. Without it, you’ll see engagement drop, product stickiness decrease, and renewals struggle.

The only real counterexample is Google, which managed a search-only interface because its addressable search universe is massive, and users come with varied needs. Even Google, however, offers dedicated pages like News or Finance to cater to specific user requirements.

If all you have on your landing page is a search box, you can expect engagement rates to keep heading one way—south!

2. Context Becomes Open Ended Without Guidance

An enterprise search box should have a business context; it’s not meant to be a free-for-all. For instance, if the AI product assists in managing CRM opportunities, an open-ended search box might invite a user to type “Where should I travel for my next holiday?” This is problematic: answering incorrectly or not answering at all creates user frustration. Instead, providing contextual prompts or “vignettes” about the recent high-value opportunities, or opportunities that require attention, will guide the users, keep them on the right path and maintain relevance.

Context can be provided in various ways—such as "suggested asks," "most asked," "top metrics," and "saved items." Any AI product should offer these options and allow users to manage them through additions, deletions, resizing, etc. This level of guidance ensures that users remain focused on tasks that drive value for the business.

3. Role-Based Authorizations Add UI Complexity

Not everyone is equal in an enterprise. Role-based authorizations are essential in enterprise environments. If not for access restrictions, many mid-market enterprises would have survived just on excel instead of very expensive CRM systems. But offering an AI search solution without permissions control could lead to legal risks. Imagine a CRM where all users, regardless of position, could access sensitive fiscal projections—this isn’t just inconvenient; it’s risky.

Implementing role-based authorizations means you’ll need several UI screens for managing users, roles, permissions, mapping authorizations, and handling connectivity to directory systems etc. Do you see the UI complexity slowly increasing now? Consider, too, how to handle unauthorized queries. If a West Coast sales rep asks, “What are East Coast sales?”—should the response say they’re not authorized, give an ambiguous reply, or show only the data that he is authorized (the user may get confused to see west coast sales when he is expecting east coast sales though)? Whether you are building a B2B product, or a B2B2C product - expect to cover corporate governance and accommodate UI design for it.

4. Not All AI Actions Can Be Addressed with Search

Assuming users will type or use voice commands for every need is shortsighted. You are not even 10% close to your destiny. Let's take an example of an AI platform trying to set up a recurring meeting: with countless configurations, toggles, and settings, achieving the same result through search or voice alone would be frustrating and time-consuming. For complex actions, users need checkboxes, toggles, and other interactive elements to reach solutions efficiently.

Remember that the answers provided by the AI platform may not be the 100% solution, and you would need to provide a way for the user to get from there to the solution. Users need follow-through actions—a pathway to refine or customize the results provided.

5. Need to Offer Collaboration Capabilities

Unless the AI recommendation is about a transient value, such as understanding a restaurant menu item, any recommendations or answers at an enterprise level would need to have collaboration capabilities to follow through across teams. Without collaborative tools, the AI platform remains incomplete.

Think of a scenario where you analyzed and identified a particular market segment that is draining your marketing resources. There can be a varying amount of follow through actions.

The typical collaboration needs would be for a group review of the findings, sharing the underlying data set, discussing the causal analysis, and talking about the impact of various options. Without providing some or all of these - you cannot build a standalone product.

6. Handling Error Scenarios & Offering Cues

When users interact with an AI product, especially through natural language, errors and misunderstandings are inevitable. A well-structured UI can provide error correction options, contextual tips, and guidance to help users refine their inputs.

Without these aids, users may become frustrated with misinterpretations, unclear outputs, or unexpected results. Visual cues and help prompts not only enhance usability but also empower users to better understand and leverage the AI's capabilities over time.

7. Customization and Personalization

Enterprise users often need to tailor their dashboards, reports, or workflows according to specific preferences or team needs. A minimalist, search-only UI can struggle to meet these demands without a robust set of customization options.

An intuitive UI can offer a layer of personalization that adapts to user preferences, enabling them to save views, set up alerts, and even organize frequently used functions for quick access. This flexibility is essential for maximizing the productivity and satisfaction of diverse user groups within an enterprise.

To sum up, the vision of a search-only UI for AI products may be compelling, but it overlooks essential elements that drive user engagement and usability in enterprise contexts. From the guiding structure of a home page to role-based authorizations, contextual prompts, and collaborative capabilities, a well-designed UI is critical for a productive and secure user experience. While a streamlined interface has its appeal, effective AI products must prioritize practical, user-centric design over minimalist trends.