Customers Don't Want Another Chatbot' - How A2UI Is Revolutionizing AI Interface

Customers Don’t Want Another Chatbot" - How A2UI Is Revolutionizing AI Interfaces

“And then we figured out that the customers didn’t want another chatbot. They neither tend to love reading nor writing that much” - paraphrasing Jakob Panten at the AWS Summit in Stockholm back in 2024.

𝗛𝗲 𝗮𝗻𝗱 𝗵𝗶𝘀 𝘁𝗲𝗮𝗺 𝗽𝗶𝘃𝗼𝘁𝗲𝗱 𝘁𝗼 𝗮 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜-𝗱𝗿𝗶𝘃𝗲𝗻 𝗽𝗿𝗼𝗱𝘂𝗰𝘁 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻 𝘁𝗮𝗯𝗹𝗲. This eliminated the need for users to craft long prompts while building concise, purpose-specific answers in the form of product comparison tables.

Written language is universal. We can express almost infinite things in written language, given we take time to think and write about specifics. This takes time.

We often help ourselves by introducing well-defined terms. Using them helps avoid re-specifying details. We can build on this abstraction and be more concise while reducing text length. This helps, but often isn’t efficient enough. That’s where we introduce specific terms and domain-specific languages.

𝗧𝗵𝗲 𝗖𝗵𝗮𝘁𝗯𝗼𝘁 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲: While language flexibility helps in building bots with wide functionality, this often requires extra work from humans writing free-form text, and might need multiple iterations to ensure all data is captured.

How did we do this before chatbots? While in Germany, you still fill out paper forms. These forms capture all relevant data. In the digital age, we use interfaces that not only capture data but provide human-friendly controls, like date-pickers, making it easier for humans to input information.

𝗘𝗻𝘁𝗲𝗿 𝗔𝟮𝗨𝗜: 𝗔 𝗚𝗮𝗺𝗲-𝗖𝗵𝗮𝗻𝗴𝗲𝗿 🔍 That’s where A2UI[1] comes into play: “Protocol for Agent-Driven Interfaces. A2UI enables AI agents to generate interactive user interfaces that render natively across web, mobile, and desktop—without executing arbitrary code.”

A2UI is created by Google and Apache 2.0 licensed. It allows agents to define interfaces on-the-fly and send them to the client. The client, with the A2UI SDK, can natively render the UI and ask users to provide the relevant data.

𝗪𝗵𝘆 𝗧𝗵𝗶𝘀 𝗠𝗮𝘁𝘁𝗲𝗿𝘀: This maintains language flexibility while allowing more specificity on required inputs and providing humans with an easy way to input data.

The protocol uses a data format efficiently created by an LLM-powered agent. It can be parsed as well. So it might be another way to have agents talk to agents?

𝗡𝗲𝗮𝘁, 𝗿𝗶𝗴𝗵𝘁?

If your head is spinning about protocols you just learned like A2A, MCP… don’t worry. They’re staying. This is complementary.

💬 𝗛𝗮𝘃𝗲 𝘆𝗼𝘂 𝘄𝗼𝗿𝗸𝗲𝗱 𝘄𝗶𝘁𝗵 𝗔𝟮𝗨𝗜 𝘆𝗲𝘁? 𝗦𝗵𝗮𝗿𝗲 𝘆𝗼𝘂𝗿 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲 𝗶𝗻 𝘁𝗵𝗲 𝗰𝗼𝗺𝗺𝗲𝗻𝘁𝘀!

🤔 𝗪𝗵𝗮𝘁 𝗶𝗻𝘁𝗲𝗿𝗳𝗮𝗰𝗲 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝗮𝗿𝗲 𝘆𝗼𝘂 𝗳𝗮𝗰𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝘆𝗼𝘂𝗿 𝗔𝗜 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀? 𝗟𝗲𝘁’𝘀 𝗱𝗶𝘀𝗰𝘂𝘀𝘀 𝗵𝗼𝘄 𝗔𝟮𝗨𝗜 𝗺𝗶𝗴𝗵𝘁 𝗵𝗲𝗹𝗽.

[1] https://a2ui.org/