From Chat to Click
Chat-based interfaces are the go-to solution for agent interactions, helping new users of agent systems navigate complex workflows. Chat is a flexible interface, and can return different formats of responses. There’s also no limit to what the user can input, if they can think of the prompt. However the user has to dream up the prompt. I think we can do better and come up with better UI/UX patterns for agent interaction using Graphical User Interface (GUI) elements. This transition mirrors a historical shift in computing itself – one that began with the invention of the GUI in the early 1980s.
What Are Agents?
In this context, agents refer to AI-driven systems built using large language models (LLMs) and frameworks like LangGraph. These agents process natural language inputs, assist with decision-making, and automate tasks within enterprise workflows. They often operate as chat-based assistants, helping users by answering queries, executing commands, and integrating with various data sources.
The First GUI: the Apple Lisa
Before graphical user interfaces became standard, computers relied on command-line interactions, requiring users to type precise text commands. This changed in 1983 with the launch of the Apple Lisa, one of the first commercial computers to feature a GUI.
Apple’s innovation drew heavily from research at Xerox PARC, where engineers had developed a visual computing environment with windows, icons, and a mouse. Steve Jobs and his team refined this concept, creating a more accessible experience for everyday users. The Lisa’s GUI allowed people to interact with software visually, using point-and-click gestures instead of memorising and typing complex commands. The GUI used the metaphor of a physical desktop and files and folders to help users understand the underlying file structure.
This shift in computing parallels today’s transition in agent interfaces. Just as the GUI made personal computing more intuitive in the 1980s, graphical UI elements can now improve user and agent interaction by reducing reliance on chat-based interactions. Instead of engaging in lengthy, text-based exchanges, agents can now use structured UI components – like dropdowns, buttons, and dashboards – to navigate tasks faster and with fewer errors.
The Limitations of Chat-Only Interfaces
While conversational interfaces have their advantages, they present several challenges:
- Efficiency Issues: Completing tasks through chat can be slow, requiring multiple messages for actions that a single click could accomplish.
- Cognitive Load: users must remember commands and context rather than relying on visual cues to guide their workflows.
- Scalability Problems: Handling multiple interactions simultaneously through chat can overwhelm agents, slowing response times and reducing productivity.
Why Graphical UI Elements Improve Agent Experience
To address these challenges, companies are integrating graphical UI components into agent workflows. This shift provides several key benefits:
- Speed & Efficiency: Dropdowns, buttons, and structured forms enable users to see what is available, and then complete tasks with fewer steps. Clicking a button is faster and easier to understand than typing out the command.
- Error Reduction: Standardised UI elements minimise the risk of typos and misinterpretations common in chat-based systems.
- Better Multitasking: Visual dashboards allow agents to manage multiple interactions without losing context.
- Improved Training & Onboarding: A well-designed graphical UI is often more intuitive than learning complex chatbot commands.
Ideas for Transitioning to a Graphical UI
If your organisation is considering moving from a chat-based agent interface to a more graphical experience, here are some best practices:
- Hybrid Approach: Maintain a balance between chat and UI elements. For example, use chat for open-ended queries while providing structured options for common workflows.
- An example of this is instead of asking the agent “what did we talk about yesterday”, you have a “History” section of your interface where previous chats and logs are stored.
- User-Centered Design: Have a strong hypothesis for what a user expects to see from an agent dashboard, and then test that with users.
- Data-Driven Decisions: Analyse chat logs to determine which interactions would benefit most from graphical UI elements.
- Iterative Rollout: Introduce new UI features gradually, test and adapt.
- Generative UI – a very experimental approach (as of March 2025) could be using the LLM to output custom interfaces for the specific scenario the user is working in, such as custom charting, displaying visual options. So the frame and window of the application stays the same, but the output section is generated dynamically on the fly.
While chat interfaces have played a vital role in agent interactions, the shift toward graphical UI elements has to happen to be a more efficient and user-friendly approach. This transition is not unlike the move from command-line computing to graphical interfaces in the 1980s – a change that revolutionised personal computing. By using graphical UI elements, we can streamline agent workflows, improve efficiency, and enhance the overall user experience.
The future of agent interfaces is visual – just as computer interaction itself has shown us.