What role should AI play in a decision system?

Uber made its name as a ride-hailing and delivery app, but it's actually a tech business that specializes in connecting people. As such, the data that informs experience improvements – such as user behavior, preferences and feedback – is central to decision-making. 

At Uber's scale, however, the utilization of AI in this process is unavoidable. And according to CX program manager Ekaterina Mironova, this creates a paradox: the more sophisticated the analysis, the easier it is to drift from the actual human experience behind it. 

At All Access: Future of Customer Insights & Data Analytics, Mironova will talk in detail about how Uber uses AI while retaining the connection with the human experience. You can register to attend the session, here.

Ahead of that, she explains how AI supports rather than replaces human judgment, why data and understanding aren't the same thing, and her three rules for deciding where AI should – and should not – be applied to customer insights. 

INLINE 

CX Network: Your session is called Beyond the hype: keeping the human in customer insights at Uber. How does Uber achieve this?

Ekaterina Mironova: At Uber's scale, artificial intelligence (AI) is essential for processing the volume of signals that come in across markets and products. But scale creates a paradox: the more sophisticated the analysis, the easier it is to drift from the actual human experience behind it. A sentiment score is useful, but it isn't the experience itself.

The way I think about our approach is to treat AI as an intelligence layer, not a replacement for judgment. It's very good at turning large volumes of unstructured feedback into patterns worth a human's attention.

The human role starts where the pattern ends: asking whether it makes sense, what's missing, what context might change the interpretation, and whether a pattern explains behavior or simply correlates with it.

The goal isn't maximum automation. It's better judgment at scale.

CX Network: How does Uber maintain human connection with its customers?

Ekaterina Mironova: Data and understanding aren't the same thing. A dashboard can tell you that a metric moved; it can't necessarily tell you what the experience meant to the customer.

That's why triangulation is so important. Quantitative data tells us where to look. Qualitative feedback helps explain why. Direct conversations surface language, context and emotion that structured data often strips out. Operational data helps us understand how widespread or consequential an issue may be.

Human connection is also about what happens after an insight is discovered. An insight sitting in a research deck doesn't change the customer experience. It has to reach the people making decisions about products, processes and operations in a form they can act on.

One of the most important roles of customer insights is to close the distance between the customer's lived experience and the people making decisions that affect it.

CX Network: When it comes to deciding where AI should and should not be applied to customer insights, what are the primary rules you abide by?

Ekaterina Mironova: I ask three questions: 

1. What task are we delegating?
2. What's the cost of being wrong?
3. And how reversible is the decision that follows?

AI is strong at scale and pattern recognition: clustering feedback, summarizing large volumes of information and surfacing emerging themes. It becomes riskier as we move from processing information to interpreting meaning, because customer feedback is highly contextual.

Two practical rules follow from that:

First, human review should scale with impact. The threshold can be relatively low when AI is helping an analyst decide where to look next, but much higher when an AI-generated insight could influence a significant product or policy decision.

Second, traceability matters. Teams should be able to move from an AI-generated conclusion back to the underlying customer evidence. Otherwise, a summary can quietly become accepted as reality without anyone questioning what was lost in the abstraction.

The question isn't simply, "can AI do this?" It's, "what role should AI play in this decision system?"

CX Network: What are some of your best practice approaches for building a VoC business case?

Ekaterina Mironova: Start with a business decision the organization is already struggling to make, not with "we need more customer feedback." 

Executives don't need more insight in the abstract; they need better evidence for prioritization.

If the priority is retention, for example, don't stop at asking what customers are unhappy about. Ask which experiences are associated with churn, which segments are most affected, and which parts of that experience are realistically addressable.
From there, I like to build a chain of evidence:

Customer signal → experience problem → customer behavior → operational consequence → financial or strategic impact.

It's also important to separate prevalence from impact. The most frequently mentioned problem isn't necessarily the most consequential one. A less common issue may disproportionately affect a valuable customer segment, create significant operational cost, or damage trust at a critical moment.

Finally, prove value incrementally. One or two high-confidence interventions with measurable outcomes can build more credibility than a large roadmap based primarily on projected value.

The strongest VoC business case isn't about proving the value of listening. It's about proving that better customer understanding leads to better business decisions

To join Ekaterina's session live on September 15, register to attend All Access: Future of Customer Insights & Data Analytics via this link

Quick links 

• How TikTok Shop cut ticket resolution time by 150%
• Inside PowerPay: Using AI agents to amplify not replace humans
• Why AI agents alone won't fix your decision-making (and 3 things that will)

40463.008 - All Access: The Future of Customer Insights & Data Analytics 2026

Join All Access: Customer Insights & Data Analytics 2026 to learn how to leverage data to gain valuable customer insights and drive business growth. 

Register Now