Implement pre-written action options in chatbot UIs

Every software vendor seems convinced that adding a chatbot to their product is the magic bullet for integrating Generative AI into the future of user experiences. We're constantly encouraged by the "talk to your application" slogans, as though conversational interfaces will revolutionise usability and enhance efficiency. Even emerging software solutions sometimes opt purely for chat-based interfaces, neglecting graphical elements entirely. But am I the only one who thinks that application designers are missing the point when they implement these chat functions?   Conversational interfaces are nothing new in any context, but since the wide spread use of LLMs, we see them everywhere. Research has repeatedly shown though, that visual interfaces outperform conversational interfaces in efficiency, task completion, and user satisfaction. For instance, a study published in Computers in Human Behavior highlighted how chatbots increase cognitive load and reduce users' sense of autonomy compared to traditional visual menus. I also remember reading another investigation at Aalborg University which found that graphical search interfaces dramatically improved user engagement and usability compared to conversational alternatives. The lack of efficiency in conversational interfaces often comes from introduction of unnecessary friction—especially when users need to quickly compare multiple data points or complete complex tasks. Visual interfaces, whether it being buttons, sliders, or other intuitive layouts, significantly reduce cognitive effort and allow users to act swiftly and confidently. So, does this mean that I don’t think that conversational interfaces have a place in our AI-driven future? Absolutely not. A great example is voice-control in vehicles or other assistive technologies, where a conversational interface clearly demonstrates its value. But these should be targeted exceptions rather than the default. So, what is the solution? I think, that instead of blindly chasing chatbot integration as a new element in our user interfaces, software developers and designers should leverage Generative AI to intelligently enhance software application capabilities without becoming fixated on the conversational interfaces as a new method of interaction. Example 1: Imagine you're working in your favourite IDE, developing a new function. The very modern IDE might offer you a Generative AI chatbot which you can ask to help generate documentation for your new function. But as clever as that sounds, wouldn't it be smarter for the publisher of the IDE to anticipate this need and simply provide an 'Export Documentation' button? The Generative AI can still handle the task of generating documentation, but as a user, you've saved yourself the cognitive effort of formulating the request. Example 2: Image you want to create a simple application which allows users to upload meeting notes for them to interact with those meeting notes in a smart way – levering Generative AI to make sense of the unstructured nature of the meeting notes. It’s tempting to create the front-end as a chatbot which simply allows the user to ask for anything they want – but as an application, it would be much more helpful, efficient and effective if the application was built to solve very specific tasks such as “extract actions”, “summarise the meeting”, etc. rather than having an open-ended set of capabilities to confuse users and make them think of how they should formulate their ask (and opening up the risk of them asking extremely poorly and getting bad outputs as a result)   I think both of these examples show how an ‘old-fashioned’ interaction method with the intelligence behind it (often referred to as Invisible AI), can cause user interactions to have fewer interruptions and no lengthy conversations at the same time as increasing quality of output — just a seamless experience where AI quietly, efficiently, effectively and invisibly works in the background. We need to shift our thinking on Generative AI from conversation to capability, leading productivity-enhancing features in a more seamless way, which in the process will remove the user as an application performance inhibitor. So let's stop asking users to talk more and instead empower them to do more. How do you see the future role of Generative AI in user interfaces?