The Case of the Addled AI

A jewelry retailer discovers the limits of automated service the hard way.

For years, Sterling & Crown had prided itself on its exceptional customer service. The family-owned jewelry store had built a reputation on personal relationships, knowing their customers by name, remembering anniversaries, and offering thoughtful recommendations. But as technology evolved, management saw an opportunity. They believed AI could enhance the business by offering 24/7 responses and handling routine inquiries, leaving associates with time for more important tasks.

Their solution was “GemBot,” a chatbot integrated across the store’s website, online chat, and text messaging. It could answer frequently asked questions, provide basic product recommendations, and even schedule appointments. Customers appreciated the convenience, and with the staff spending more time on in-store clients, it seemed like a win-win.

Each year, Sterling & Crown’s in-store custom design event drew a surge of customers. Among them was Julia, a longtime client who had been shopping with the jeweler for over a decade. Her purchases were always significant custom pieces and high-ticket designs requiring expert attention. This year, she had something extra special in mind.

Julia wanted a custom-designed diamond necklace for her daughter’s wedding. She envisioned a 5-carat diamond centerpiece, incorporating sapphires from her late mother’s jewelry collection — a sentimental touch honoring the bride’s great-grandmother. Her budget was $12,000, and she trusted Ava, her go-to salesperson, to bring her vision to life.

Late one evening, Julia reached out through the store’s website, expecting to hear from someone first thing in the morning. Instead, she got GemBot.

She detailed her request through the site, expecting a confirmation for an appointment with Ava. Instead, the chatbot responded with generic diamond pendants from the store’s online catalog. No appointment was scheduled, and no human followed up. Frustrated, Julia sent another message, trying different phrasing, hoping to bypass the bot. The responses remained seemingly canned and unhelpful. Though irritated, Julia decided to call the store directly the next day.

By morning, Sterling & Crown was overwhelmed with foot traffic from their annual event. Staff barely had time to check the phones, assuming GemBot was handling inquiries.

When Julia finally got through to a real person, she was told Ava was with another client. The associate quickly jotted down her name and number, instructing her to schedule online, the same system that had already failed her. Feeling dismissed, Julia decided it wasn’t worth the hassle of continuing to try and reach out.

Her message was placed near the phone. And forgotten. Days later, after the event ended, the associate finally mentioned the missed message to Ava. Realizing the mistake, Ava called Julia immediately. But it was too late.

Julia had already taken her business elsewhere. Another jeweler had responded promptly, worked within her budget, and created the necklace in time for her daughter’s wedding. In addition, the competitor also upsold her three solitaire necklaces for the bridesmaids, a $2,000 add-on sale that Ava never had the chance to make.

Everyone at the store had assumed GemBot was working efficiently, but they had failed to set up safeguards. No one checked on the automated messages to follow up on abandoned conversations. In their eagerness to embrace AI, Sterling & Crown had seemingly lost sight of what had made them successful: personal service.

Ultimately, it had cost them a loyal, high-value customer. Now the owners had to ask themselves: Had they gone too far with AI, or had they simply failed to use it correctly? And how many other clients quietly slipped away, unnoticed?