in this issue
|
A real-world look at how Uber is using AI to uncover hidden customer signals, improve support quality, and build a more proactive digital experience on a global scale.
|
by Anindya Sundar Das, Senior Director, Head of Global Digital Experience at Uber
Uber’s Global Digital Experience team deployed an AI engine that infers customer satisfaction across every support interaction, not just the small fraction who complete surveys. Structured around three core metrics (Resolution, Effort, Sentiment), the model transforms scattered interaction data into a real-time quality signal, enabling proactive improvement at a global scale.
Digital Experience at the Center of Support Strategy
The Global Digital Experience team serves as the strategic bridge between Customer Support Operations and Tech teams. We operate at the intersection of innovation and execution, ensuring that our support technology stack is not only world-class but purpose-built to drive both a superior customer journey and operational excellence.
Our mandate is twofold:
- We collaborate directly with Product teams to architect high-impact solutions that elevate the user experience and enhance support efficiency.
- We oversee the end-to-end deployment of these technologies across all global markets. By aligning product capabilities with local market needs, we ensure we are maximizing our ROI and capturing the full commercial value of our Digital investments.
In short, we transform technological potential into exceptional customer support experiences on a global scale.
The Feedback
Historically, we were limited by the “feedback gap” — only seeing customer satisfaction feedback from a small percentage of customers who chose to fill out a feedback survey after the actual interaction. To support operational excellence, we have launched an AI engine that analyzes every customer interaction in real time to help predict customer satisfaction with the interactions and uncover hidden insights. It’s a new way to evaluate the quality of our customer support.
An AI Model Built on Three
Our new AI model solves this by inferring satisfaction scores across all the interactions, effectively unveiling deep-tier insights even when a customer provides no explicit feedback. By proactively identifying where an experience would have fallen short of our standards, we are no longer just reacting to problems; we now have the ability to capture the intelligence necessary to take corrective action and systematically engineer a superior support experience for the future.
Our strategic approach to these interactions is built on three pillars: Resolution, Effort, and Sentiment. We designed the AI model around these three pillars:
|
RESOLUTION
This is our primary goal. We are solving the problem the customer approached us for. We measure our success by our ability to definitively settle the customer’s issue and make things right.
|
EFFORT
Getting the resolution should be easy for the customer. We leverage our AI model to quantify the customer’s effort score, ensuring the path to a solution is as frictionless and intuitive as possible.
|
SENTIMENT
Our goal is to create a positive delta. We analyze the shift in customer emotion from the start of an interaction to its conclusion, ensuring they leave the experience more confident in our brand than when they arrived.
|
To achieve this, we synthesize a lot of data: examples include transactional telemetry (real-time trip and order data) and conversational intelligence (interaction logs, turn count, and tone of expression). We gain a holistic view of the customer’s needs by combining these. The deep integration of these data points allows us to make informed, empathetic decisions at scale, transforming every support touchpoint into a driver of long-term value.
Cross-Functional Build, Hard-Won Lessons
This initiative required strategic alignment and deep, iterative collaboration across various parts of the organization to ensure it translates into something usable and actionable. We partnered extensively with our Data Science and Analytics teams to define Resolution, Effort, and Sentiment. This wasn’t just about math; it was about ensuring our definitions captured the nuanced insights necessary to empower our global operations.
There were indeed many learnings. While we entered the initiative having dealt with years of CSAT data, we quickly discovered that traditional surveys only scratch the surface of true customer sentiment. As we peeled back the layers to define more sophisticated metrics and sub-metrics, we uncovered a level of nuance that our legacy CSAT simply couldn’t capture. Each iteration didn’t just move the needle; it fundamentally hardened our infrastructure and made our solutions more robust.
Aligning diverse stakeholders on a single, “composite” definition of customer satisfaction required a lot of iterations. We debated how disparate data points, from resolution speed to emotional tone, should be weighted and integrated into a meaningful KPI. In an environment as vast as ours, spanning multiple business lines, languages, and global markets, “evaluation” was incredibly demanding. Teaching the AI not just that it failed, but why it failed within a specific cultural or operational context, required a lot of effort.
Every challenge we’ve encountered has served as a step for improvement. We are emerging from this process with a more sophisticated understanding of our customers and a more resilient platform. In the world of AI, the friction of today is the fuel for the breakthroughs of tomorrow.
What Changed and What’s Ahead
This initiative is a critical catalyst for our next phase of growth, delivering two distinct benefits:
| 1 | We are moving beyond anecdotal feedback to a robust, data-driven engine. We can now identify systemic friction points across our product interactions and support policies with unprecedented scale. This allows us to prioritize high-impact optimizations that move the needle for our customers. |
| 2 | As we pivot from the world of legacy RPA to a dynamic, conversational AI ecosystem, we solved a fundamental challenge of comparability. This intuitive insights layer allows us to normalize performance metrics across diverse technologies. For the first time, we can effectively “compare apples to oranges,” giving us a unified view of customer support efficacy regardless of the underlying tech. |
In effect, this initiative accelerates our velocity of innovation, helping us deploy solutions faster and with higher confidence, ensuring that every advancement in our support technology translates directly into a more seamless, superior experience for our customers. AI is transitioning from a supporting tool to becoming the bedrock of our customer support systems. At our scale, AI provides a level of operational sophistication that was previously unattainable. It is not constrained by scope related to geography, product or issue categories, or even languages. AI models can ingest disparate, high-volume data streams—stitching together transactional history, real-time telemetry, and natural language conversations—to render complex decisions quickly. We are effectively transforming scattered data into seamless customer journeys. The AI architecture provides superior depth in decision logic. It can articulate the why behind a resolution with greater detail than human-led processes.
What’s Next for Digital Experience at Global Scale
 |
In the medium term, AI should be able to remove the traditional trade-offs between scale and personalization, creating a possibility of leveraging AI to deliver a gold-standard experience for every user, on every continent, every time. However, the AI trajectory points toward a paradigm shift in the longer-term future in how we serve our customers.
There are at least three ways in which this could manifest:
The technology is in a phase of rapid evolution, but the insights we gather through experimentation today will define the market-leading solutions of tomorrow. The applications for AI in the CX context can be many. It can be deployed to identify opportunities and coach human customer support agents in real time, to make high-integrity independent decisions about customer problems, to audit decision accuracy and predict customer satisfaction, or to identify root-cause improvements to build a more robust, future-proof support architecture. |
| KEY TAKE AWAY
This is a frontier technology, evolving at a rapid pace. Evaluating, training, and deploying these models is non-trivial. However, AI is much like a high-potential prodigy. It requires rigorous attention, structured support, and constant refinement. If we commit to nurturing this capability with the right resources and focus, it will mature into a competitive advantage that transforms our businesses. The goal isn’t to be perfect on day one; it’s to be better every single day thereafter. |
 |
About the Author
Anindya Sundar Das is a global business and technology leader with over 20 years of experience driving operational transformation and AI-driven innovation across multiple continents. Currently serving as Senior Director and Head of Global Digital Experience at Uber, Anindya is a recognized pioneer in automation and Generative AI, having delivered massive impact through globalization and scaling of advanced Gen AI and automation platforms. Recognized in the “Who’s Who in the U.S. in Tech,” he has previously been COO at CK Birla Healthcare and Business Head at Orient Electric. A double-medalist and top-ranked scholar with a career marked by prestigious accolades, he was formerly Associate Principal at McKinsey & Co., leading major transformations for global automotive and mining organizations. |
