Dear AI, We Need to Talk: Hundreds of Retail Pros Say What They Really Think
We didn’t set up the typical booth at this year’s Shoptalk. Instead, we brought an artificial intelligence confessional.
On one side was a wall of color‑coded sticky notes tied to familiar retail customer experience pain points. On the other side was a simple prompt that disarmed people as they walked by: “Dear AI, we need to talk. What went wrong?”
We designed the wall around five familiar CX themes:
- Customers repeating themselves (no shared context across channels or past conversations).
- Moments of friction in the journey (too many steps, clunky handoffs, long waits).
- Missed opportunities for personalization (generic offers, ignored loyalty, off‑base recommendations).
- When AI gets it wrong (tone‑deaf, canned answers to sensitive issues).
- “Dear AI, here’s what I actually want” (what good looks like from a customer’s perspective).
It's clear from the themes we presented that we knew attendees would document examples of automation falling short of what shoppers actually need from brand support and shopping journeys. But still, we were surprised by what we learned.
Once attendees emptied out their customer experience stories, people started writing as AI users themselves. Their frustrations lined up with those of their customers. The problems they talk about fixing for their customers are the same problems they run into when they use AI at work. Across that wall, six themes emerged.
6 Ways Retail Pros Are Frustrated With AI
1. Trust and Accuracy
"Can’t trust data,” “pulling false info,” and “need to verify sources” were among the responses that blanketed the wall. If a user's default behavior is opening a second tab to verify what they've just been fed, that tool quickly shifts from resource to suggestion, not a decision partner or well of trusted information. That hidden verification work only grows as decisions get more important.
2. Tone Problems
The second cluster read more like relationship feedback than software feedback.
- “Too positive.”
- “Way too positive. Be real.”
- “You say all my ideas are great. They are not.”
- “Not being candid enough when I’m looking for constructive feedback.”
These reactions show us that what AI says is just as important as how it says it. Responses that are always agreeable and overly apologetic are dishonest and indirect. Retail leaders and their customers want authentic interactions, not the generic and overscripted content that we all experience.
3. Output Quality
The third bucket focused on the small problems that add up in a professional context.
- “Bad graphics and spelling in images.”
- “Learn to spell in images.”
- “Too many commas.”
- “Quantity over quality responses.”
- “Verbose, not to the point.”
Designers and marketers described losing time cleaning up AI‑generated images with spelling errors, trimming overformatted outputs into something that matched their brand, and cutting through responses that felt too loud and too complex for the task at hand. They didn't want to spend their day fixing punctuation, stripping out heavy formatting, or wrestling with walls of text to find one useful insight.
4. It Creates New Work
If you’re already seeing issues with trust, accuracy, tone and output quality, then the consequences outlined in theme No. 4 shouldn’t come as a surprise. Leaders expect AI to make their lives easier. Instead, they end up working more and losing time in the process.
Here are some notes folks who came to our booth jotted down:
- “Have to repeat myself!”
- “Requires multiple threads/convos to get an accurate answer”
- “Takes too long to get the right outcome”
- “Too slow!”
- “Usage limit hit”
5. The Context Gap
The fifth bucket of notes went deeper into why people felt let down by AI.
- “It doesn’t have the context in my brain.”
- “Too much confidence, too little context.”
- “Bad direction.”
Without real memory of the broader relationship, the brand, the task at hand, or the prior conversation, models make guesses. And they do so with unwarranted confidence. Sometimes the AI guesses well, but many times it does not.
On the wall, we saw feedback calling out AI that doesn’t understand someone’s face, lacks uniqueness or accuracy, or, paradoxically, feels like it knows them too well in ways that come across as unsettling.
6. The Bigger Picture
Finally, there was a more existential row of notes.
- “People are losing their ability to solve problems themselves.”
- “Homogenization.”
- “Loss of connection with consumers.”
- “The laziness of it all.”
- “Feels like cheating.”
Underneath the practical complaints about hallucinations, tone, and cleanup, many people worry that AI is making content more of the same and thinking less independently. The fear was not only about poor outputs today, but about what happens if organizations normalize them over time: "Everyone has it but doesn't know what to do with it."
What This Means for Retail
Our wall gave us a clear view into how interacting with AI today feels from both sides of devices: The same themes that frustrate brands mirror bad customer experiences.
Therefore, brands powered by AI have to contend with this uncomfortable question: If these problems are so widely recognized, why do they still show up everywhere in how AI is built and deployed?
Why These Issues Exist and Persist
Several structural forces help explain that mismatch. The incentives veer toward speed, coverage, and risk mitigation. They rarely focus on the quality of interactions in a way that a customer would find valuable. That leaves brands with generic products that don't align with its history and voice.
There's also a “too many options, not enough depth” problem, something a Shoptalk attendee called out. With so many AI products in the market, it’s easy for brands to choose platforms that do the headline tasks well while quietly trading off the granular details that actually define how an interaction feels. On paper, those decisions look rational, but in practice they can produce yet another bot that technically works but still misses the point.
Finally, there's a governance gap. In many organizations, no one is explicitly accountable for questions like:
- Does this brand’s AI sound like the brand?
- Is it honest about what it cannot do?
- Does it reduce work for customers and agents or simply reshuffle it?
Absent that ownership, these frustrations with AI that show up on sticky notes remain personal and tactical instead of becoming lessons applied to how AI is selected and implemented.
Dear AI, Here’s What We Actually Want
There's one more column on the wall we haven't talked about yet. Under “Dear AI, here’s what I actually want,” Shoptalk attendees largely didn't ask for models that spoke better or flashier demos. They asked for less friction. Those notes read like a brief for what brands need to focus on next.
It starts with implementing these sticky‑note frustrations as requirements for something better. Brands must look at AI projects through a different lens: not only whether they can answer more questions or automate more steps, but whether they avoid pushing extra work and doubt onto the people on the other side of these tools. It means evaluating AI vendors on how they handle context retention, tone, and trust. And it also means being deliberate and transparent about where AI is placed in the journey, when it hands off to a human, and how it’s monitored over time.
The wall at Shoptalk did more than surface industry grievances. It offered a set of criteria for what comes next in our AI adoption journey. When evaluating what to build, buy, and put in front of customers, brands must apply this criteria if they want positive and long-lasting relationships with their customers.
Charles Studt is chief marketing officer of Delight.ai, a smart AI concierge that turns every customer moment into recognition, loyalty, and growth.
Related story: What’s the Missing Link in Consumer AI Agents?
- Categories:
- Artificial Intelligence (AI)
Charles Studt is the chief marketing officer of Delight.ai, the AI-powered customer concierge platform that helps global enterprises deliver predictive, proactive and personalized customer experiences across every channel.
With more than 20 years of experience scaling high-growth technology companies and building category-defining brands, he leads Delight.ai’s global marketing strategy at a pivotal moment in the evolution of agentic AI and customer engagement. As CMO, Studt is responsible for strengthening Delight.ai’s position as a leader in agentic AI-powered customer experience, bringing the company’s AI innovations to new markets and showcasing the measurable business impact of AI-driven customer engagement.
Before joining Delight.ai, Studt held executive marketing roles at Qualtrics, where he led growth for their Customer Experience platform, cementing it as the leading CX solution for global brands. He has also held marketing leadership roles at IntelePeer, QuanticMind and Redbooth.





