Uber and OpenAI: A Partnership to Revolutionize Urban Mobility

Uber and OpenAI: A Partnership to Revolutionize Urban Mobility

Uber, the giant of urban mobility, has recently strengthened its partnership with OpenAI to integrate intelligent assistants and voice features into its services. This collaboration aims to enhance the user experience for both drivers and passengers by leveraging cutting-edge technologies to optimize operations on a global scale.

Every day, millions of users turn to Uber for various services, ranging from ride bookings to food deliveries. Behind these interactions lies a complex system that must adapt in real-time to variables such as traffic, weather, and local events. Uber manages 40 million rides daily with 10 million drivers and couriers spread across 15,000 cities in over 70 countries. Each city presents its own challenges, requiring continuous and precise adaptation.

Since its inception, Uber has utilized machine learning to support its infrastructure. Today, with OpenAI's advanced language models, Uber can process complex data more quickly and provide seamless conversational responses, thereby enriching the user experience.

The collaboration between Uber and OpenAI enables the development of innovative products that simplify earning opportunities for drivers while reducing friction for passengers. With these models, Uber can now deploy solutions more quickly and efficiently than ever before.

“For the first time, technology drives what can be solved. Problems that once seemed out of reach can now be addressed.” — Aarathi Vidyasagar, VP of Engineering and Science

Transforming Complex Market Data into Real-Time Advice for Drivers

For Uber drivers, flexibility is a major asset. Some drive full-time, others only on weekends, or in between classes. This flexibility requires constant evaluation of opportunities, raising questions such as:

• Where should I position myself right now?
• Is it worth driving to the airport?
• Should I switch from rides to deliveries during lunch?
• Why do my earnings seem different today?

To address these questions, Uber has developed Uber Assistant, an AI assistant designed to support drivers throughout their journey on the platform, from onboarding to daily earnings optimization.

“We want to empower drivers to make better decisions by providing them with a summarized view of the market and real-time insights,” says Dharmin Parikh, Director of Product Management at Uber.

The assistant helps drivers maximize their earnings by transforming complex data like earnings trends and heat maps into practical advice. Drivers can ask questions in natural language and receive personalized responses while navigating the app.

Uber's goal is to reduce cognitive load, meaning the effort required to interpret complex data while trying to maximize earnings. This approach has proven particularly beneficial for new drivers, who can learn market dynamics more quickly.

Although initially designed for new drivers, Uber Assistant is also used by experienced drivers, who find it a valuable tool for optimizing their time on the platform.

“The Assistant helps drivers get up to speed quickly, compared to the need to complete hundreds of rides to understand how the platform works,” explains Parikh.

Building Trust at Scale with a Multi-Agent AI System

For Uber, accuracy, safety, reliability, and speed are priorities when implementing AI systems. Considerations include adherence to policies and latency that meets user expectations.

Uber has designed Uber Assistant around three core principles: safety, trust, and low latency. Uber's multi-agent architecture routes each user request to the most appropriate system, ensuring that every query is handled with the necessary attention.

For simple tasks, Uber uses fast models, while for complex tasks, more advanced reasoning models are employed. Uber has also developed AI Guard, an internal governance layer that filters prompts and responses to ensure safety and consistency.

When drivers receive accurate recommendations, they return, ask more questions, and spend more productive time on the platform.

“If users don’t trust the system, you lose them quickly,” says Parikh. “But when they see value, they come back.”

Expanding Accessibility with Voice

Uber is also integrating OpenAI's Realtime APIs to develop voice features, a major shift in the user interface.

While typing in an app is efficient for simple requests, many transportation needs are more complex. For example, a user might want to say: “I have five bags and five other people with me. I need a nice car to go to the airport. What do you recommend?”

Uber's new voice features allow users to request a ride in natural language. The system uses the Realtime API to interpret intent and make recommendations, while synchronizing verbal and visual responses in the app.

This could mean suggesting UberXL for rides with a lot of luggage or recognizing saved destinations.

“Voice removes the barrier of accomplishing one task at a time,” says Parikh. “You can express your complete intention naturally, and the system can orchestrate the outcome.”

Voice also enhances accessibility and unlocks new workflows. For drivers, it allows interaction with the app hands-free, and for passengers, it simplifies interactions.

“Voice removes the barrier of multi-tap because you can say multiple things,” says Vidyasagar. “This unlocks the ability to connect different parts of the ecosystem.”

Note: The voice booking feature will be rolled out in the coming weeks.

Faster Iterations, Stronger Teams, Better Products

With the rapid evolution of LLMs, Uber has changed how it builds its teams. Engineers work with retrieval systems, evaluation pipelines, and orchestration frameworks. Product, legal, operations, and design teams collaborate to define policies and improve user experiences.

Intelligence is no longer confined to a centralized team but is integrated throughout the company, accelerating experimentation and innovation.

“It’s no longer a specialized group doing all of this,” says Vidyasagar. “Many teams can contribute because the barriers to building have been lowered.”

This shift fosters experimentation and generates new ideas within Uber's ecosystem.

“Every ride, every trip is a sequence of events, and understanding and processing that nuance is what the LLM unlocks for us,” says Vidyasagar. “It gives us a lot of information about the direction to take, and this unlocking — at the scale we have — is exceptionally powerful.”

Expanding Intelligence Across the Market

Uber Assistant is now being deployed across the driver network in the United States as part of a pilot project, while Uber continues to test and refine the experience:

• Hundreds of thousands of American drivers have access to beta experiences of Uber Assistant.
• Support for drivers at the beginning of their journey is enhanced, helping new drivers position themselves better for more rides.
• Strong and repetitive engagement, with users returning after successful interactions.
• Better use of time on the platform through smarter market insights.
• Faster product iteration cycles due to model specialization and continuous evaluation systems.

From helping a new driver get their first ride to optimizing earning opportunities for an experienced driver, Uber is using OpenAI's models to make work more productive, transportation smoother, and daily logistics more human.

“As an engineer, OpenAI simply unlocks the ability to solve these problems in a different and unique way,” says Vidyasagar.