United Rentals

AI Voice Assistant

United Rentals is the largest equipment rental company in the world. The company’s approximately 26,000 employees serve construction and industrial customers, utilities, municipalities, homeowners and others.

Business Challenge

An AI-driven assistant app “AI Max” that enhances engagement through voice-to-voice experience and automates responses to support a salesperson placing a complex order while on the move.

What we did

Fully functional application on web, Android and iOS

Designed to place a draft reservation in their UR Max saved cart with draft information using voice interaction.

A custom voice assistant

to place orders leveraging 1000 of equipments, customer and jobsites with unparalleled speed and accuracy.

Speech To Text (STT) and Text to Speech (TTS)

Using OpenAI’s whisper, and have the ability to switch to groq within the application.

Fully functional application on web, Android and iOS

Designed to place a draft reservation in their UR Max saved cart with draft information using voice interaction.

A custom voice assistant

to place orders leveraging 1000 of equipments, customer and jobsites with unparalleled speed and accuracy.

Speech To Text (STT) and Text to Speech (TTS)

Using OpenAI’s whisper, and have the ability to switch to groq within the application.

How we did it

Custom AI Agent Solution

At the start, we aimed to build a fully customizable AI system using Autogen to orchestrate a group of specialized agents.

Agents as a Service

To reduce complexity and expedite development, we transitioned to AWS Bedrock Agents, a managed ecosystem designed to leverage pre-configured agent architectures. Bedrock offered out-of-the-box support for multiple Large Language Models (LLMs), including Claude, Llama, and Mistral.

Improving Speed and Tool Calls

To enhance performance, we revisited the architecture, focusing on optimizing tool usage and response times.

Delivering a working prototype

The prototype proved the concept but highlighted issues with accuracy, customization, and audio quality. Iterative improvements in STT/TTS flexibility, state management, and user feedback created a faster, user-friendly experience.

Key features Delivered

Voice Calibration

Calibrate your voice to help Max accurately recognize your voice in your current environment.

Assigned customers

You can find all of your assigned customers along with the favorites you chose in UR max.

Change speech models

If you are not happy with the way your voice messages are trascribed to text, you can try changing the model.