The Solution:
Due to the large disparity in potential broker users and the potential need for deep navigation, importance was placed on the AI assistant interface to simplify many complex tasks. This was for the following reasons:
- The “Broker OS” AI assistant can bridge the discoverability gap for all user personas with minimal training of an LLM. With a handful of flexible response templates, the UI can easily accommodate many complicated user requests.
- Users with limited time or limited tech know-how can easily “talk to” the AI assistant to complete many common tasks using natural language.
- The AI assistant can utilize user behavior data to provide contextual suggestions based on the type of user and where in the app the user initiates the assistant UI.
Most input commands to the AI fall into one of two categories:
a. Task-based commands (Make an appointment, set a reminder, take a note, call a contact). Each of these have a corresponding manual workflow, however a properly formed command (ex: “set an appointment with Jim Roberts on May 10th at 10:00 a.m. for 30 minutes“) to the AI can fast-track the whole process.
b. Visualization-based requests (who are my top performing agents, How are my offices tracking towards their goals, etc.). Each of these have a manual flow to access related visualizations.
Example Screens:
