I like that the interface of your agent is different than ChatGPT. It’s light in all aspects of the word. That said, I mentioned potential improvements in user flows that might make the overall experience even smoother than it is now. That’s why I described a few raw improvement ideas, some low-hanging-fruits:

• Place a dataset selection button inside the request field instead of the side panel, because:

  • There is no way a new user will know about the side panel

  • Placing a database selector inside the request field automatically highlights the next step a user should do

    

• Move dataset selection options into a dropdown menu from the side panel to:

  • Not scare users with large side panels that might be perceived as switching content.

  • Keep a user's focus point at the same location.

  • Of course, there will be less space, but a few simple optimizations will allow you to minimize this negative effect.

    

• Add a dataset exploration button for access to get access the dataset management part of the platform, so:

  • Users will know about the dataset management part of the system

  • Navigation from chat. → app. will be clear

    

• Update dataset search’s work logic:

  • Search through only the user's own datasets. Because it’s impossible to understand what a found dataset contains when only its name and date are visible.
  • Hide search if a user has <10 datasets. This will save space and reduce cognitive load for new users.

As a result, the user flow could look like this:

There is also a problem with the pricing calculator: nothing changes except price:

<aside> 👨‍💻

These are only a few things that I mentioned when tried the system. I’ll gladly dig much deeper into the system and user flows, so the platform could be even more amazing than it is now. Feel free to drop me a line [email protected]

Anton, a product designer with 9+ YoE in product design, and a background in engineering and analysis

</aside>