How In-the-Moment AI-Video Interviews Can Turn CX Metrics into Diagnoses

1. Background

For quick service restaurants (QSR) CX measurement is of strategic significance. Major chains like McDonald’s, Starbucks, Yum! Brands, Domino’s invest heavily in digital ordering platforms, loyalty apps, AI-driven personalization, VoC analytics, and CX measurement tools because speed, convenience, reliability and satisfaction are core to their value propositions. 

The high frequency of daily consumer interactions creates massive volumes of feedback data. However, (QSR) customer experience measurement and VOC programs have optimized for speed and scale. Short post-visit surveys, large samples, and tightly tracked KPIs such as CSAT, NPS, and perceived speed form the backbone of most CX dashboards.

Yet many insights and CX leaders recognize a persistent gap: they can see the numbers and their movement, but struggle to understand the context and explain why. A six-point drop in “speed satisfaction” would raise urgent questions but CX surveys would rarely tell whether the issue lies in staffing, queue design, menu clarity, order staging, or communication breakdowns. The result is a familiar pattern across service industries: dashboards without diagnosis.

This article argues that complementing such measurements with in-the-moment, AI-moderated interviews offers a practical way to close that gap – without sacrificing speed or scalability. 

2. The CX Measurement Trap

It is hard to make CX measurements based on quantitative surveys actionable, yet they have a big influence, and managerial actions are taken on them. The following issues create this tricky situation.

The limits of a 90-second survey. QSR CX surveys are short for good reasons: response rates matter, friction and interruption must be minimal, and programs often run continuously. But brevity comes at a loss. Most surveys deliver high-level evaluations (“satisfied,” “dissatisfied”) about broad constructs (“speed,” “cleanliness,” “service”) but are sparse in context on what actually happened during the visit. This is misaligned with how QSR experiences succeed or fail for people: through micro-moments of truth, often shaped by situational factors such as peak-hour congestion, unclear pickup flows, staff interactions or kiosk usability issues.

Volume does not equal understanding. Large datasets create confidence (or a perception thereof), but not necessarily clarity. When CX metrics move, teams are left debating competing explanations: execution vs. process design, staff behavior vs. environment, or service vs. product quality. As service researchers have long noted, measurement without explanation risks optimizing the metric rather than fixing the experience (Grönroos, 1984; Johnston & Clark, 2008).

3. Our Method: Capturing Live Uninhibited Qsr Experiences through AI-Moderated Interviews

To help overcome these caveats and add richness to CX measurement, we ran 75 AI-moderated video interviews through Conveo (www.conveo.ai). The interviews lasted ~15 minutes each and were conducted across 28 QSR brands, completed in less than three days of fieldwork. All interviews were immediately transcribed and coded within the Conveo AI platform.

The interview protocol was as follows. Participants visited a QSR of their choice, with half of our sample that went to a brand they regularly visit and the other half to one they rarely or never visit. The interviews started immediately after ordering, continuing through receipt and first consumption to capture the full-service sequence in the moment.

These AI-led interviews capture experiences while fresh in memory and judgments are vivid, people are in the consumption context, displaying concrete tangible cues (e.g. menu boards, packaging …) and showing real emotions tied to specific moments of friction or delight. 

An additional advantage is that our approach does not suffer from the so-called Hawthorne effect - the phenomenon that behaviour changes due to the mere fact that the subject knows their behaviour is being observed often due to human presence. Our AI moderator captures truly uninhibited customer experiences.

With Conveo’s AI moderation all of this is feasible at scale, combining structured probing with natural storytelling – something traditional qualitative methods struggle to do quickly and cost-effectively.

4. Anchoring Our Findings in Servqual: Still Relevant, Often Underserved

An advantage of Conveo’s GenAI platform is that brands can instruct the agent to capture, anchor and analyze the unstructured interview information using any conceptual framework without overtly probing for it during the interviews. Contrary to surveys which often follow structured models, this leaves room for natural spontaneous conversations and capturing experiences as they unfold. 

We applied the SERVQUAL model (originally developed by Parasuraman, Zeithaml, and Berry (1988)). The SERVQUAL remains one of the most practical lenses for diagnosing service quality. It decomposes customer perceptions into five dimensions:

• Reliability – delivering what was promised accurately and consistently

• Responsiveness – willingness and ability to help promptly

• Assurance – competence, confidence, and trustworthiness

• Empathy – warmth, care, and individualized attention

• Tangibles – physical evidence of service (environment, signage, equipment)

In QSR contexts, these dimensions are anything but abstract. They map directly to lived judgments: Did I get what I ordered? How long did it take? Did the staff seem competent? Did they treat me well? Was the restaurant clean and easy to navigate?

5. What Did We Learn?

Our platform automatically structures the data and generates real-life video quotes, rich summaries and charts. One of the most powerful and managerially relevant features, however, is our “Talk Your Data” module. This capability allows prompting your interview data to solve specific management problems and question. For this article, we prompted the AI agent to analyze our interviews and assess the performance of QSRs along the SERVQUAL dimensions and assess how each dimension affects customer satisfaction.

What Customers Actually Talk About

Across the ordering and consumption process, visitors of QSRs talk the most about responsiveness, tangibles and reliability (over 1/5 service quality mentions) and to a lesser degree about empathy and assurance. Our data do show underlying differences when it comes to driving satisfaction, however. Reliability and assurance are must-have basics, but responsiveness, tangibles and empathy make a true difference in satisfaction.

 

 

What Sets Things Apart

Responsiveness is the hero metric but also the primary swing factor. Speed dominates QSR evaluations, but not as a stopwatch. Customers experience responsiveness through progress visibility. Long waits with no explanation or unclear pickup cues create disproportionate frustration (the “why is this taking so long?” effect), even when actual wait times are moderate. Improvements in responsiveness often come as much from communication and design cues as from additional labor.

“It seemed like it was a dead space. I didn’t know what was really going on and it’s hard to know where you’re at in the queue … There’s no acknowledgement… which order they’re working on.”

“I was actually the only person… and it still took forever. I bet I waited for over ten minutes… he just didn’t seem to be in any hurry whatsoever, and I was getting frustrated standing there and waiting.” 

Tangibles are silent when good, destructive when bad. Cleanliness, layout, and menu clarity were rarely praised, but strongly condemned when lacking. Dirty tables, messy kiosks, wet floors, overflowing towels in restrooms are recurring weaknesses that create sharp negative spikes. Tangibles function as trust signals, influencing perceived food safety and operational competence.

“It’s winter right now, so there was lots of water getting tracked in… The floors are very dirty.”

“The restroom was mildly dirty. The paper towels were kind of overflowing in the waste disposal area, and the toilet wasn’t the most clean.”

Empathy makes the difference between “just fine” and “very satisfied”. Empathy can be delighter even though people do not expect personalization in a QSR. When warmth or courtesy stood out, satisfaction was elevated. When warmth / courtesy becomes the standout, it tends to elevate an experience from acceptable to memorable - especially in a category where functional delivery is expected. Empathy also acted as a buffer when speed faltered – supporting prior research showing that interpersonal treatment shapes forgiveness and recovery perceptions (Smith et al., 1999). 

“My cashier person was super nice… radiant. And then what I really liked… they wrote ‘you look cute today’ on my cup of coffee, and I just thought it was the nicest thing ever.”

“There’s nothing worse when you go into a place and you’re not acknowledged at all… They don’t say, ‘we’ll be with you in just a second’… The least you can do is acknowledge my existence when I come in.” 

6. Conclusion

Our approach changes the CX playbook!

Surveys excel at trend detection, benchmarking, and early warning signals. Rapid live AI-led interviews reveal how service breakdowns manifest in real operations: unclear menu cycles, ambiguous pickup flows, unacknowledged waits, or late-discovered fulfillment errors. Together, they form a measurement and diagnosis system: a structure long advocated in service management but rarely operationalized at speed and scale.

If insights teams start combining surveys for monitoring and in-the-moment interviews for diagnosis they can move faster, closer, and more diagnostically. These insight teams can then stop being reporters of CX outcomes and start becoming partners in fixing the experience itself.

Note. We have much more concrete findings to share. If you are interested in more, please contact the authors.

References

• Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). SERVQUAL: A multiple-item scale for measuring consumer perceptions of service quality. Journal of Retailing.

• Grönroos, C. (1984). A service quality model and its marketing implications. European Journal of Marketing.

• Smith, A. K., Bolton, R. N., & Wagner, J. (1999). A model of customer satisfaction with service encounters involving failure and recovery. Journal of Marketing.