Let’s talk about my recent experiences using AI in customer service.
Like many Americans, I bought a new TV last week — not for a Super Bowl party, but because the ice storm in middle Tennessee caused a power surge that took out our living room TV. (Side note: make sure you’re using a surge protector for your TVs… and check the expiration date. Apparently those go bad after three to five years. Who knew?!)
Standing in the store, overwhelmed by options, I did what I increasingly do when I’m unsure: I recorded a quick video of the TVs, uploaded it into ChatGPT, and asked which one to buy. Five minutes later, without ever speaking to a single store employee (scan-and-go for the win!), we were on our way home with the TV AI recommended.
It was a great example of how technology can remove friction from decision-making. A few days later, I had a very different experience with AI in customer service while trying to fix a delivery mistake — and the contrast was striking.
Customers Don’t Love AI — They Love Easy
Every week, there’s a new think piece about AI and customer service. AI tools were even the subject of multiple Super Bowl ads this year.
There’s so much dueling noise claiming, “Customers HATE AI!” and “Customers LOVE AI!” Here’s the truth: customers don’t love or hate AI. They love easy and hate hard.
Whether you’re debating chatbots, automation, or other forms of AI in customer service, the bar customers use to judge the experience is surprisingly simple: did this make things easier for me, or harder?
It’s never been about the tool itself; it’s about how the tool is implemented. This is especially true when it comes to post-purchase experiences, where AI is often positioned as the future of customer support and service recovery.
Three of my own recent post-purchase experiences perfectly illustrate the difference between AI in customer service done right, done poorly, and done intentionally.
Three Real AI in Customer Service Examples
When AI Nails the Service Recovery
In December, I ordered four adorable Peanuts stockings from PersonalCreations.com. Three were perfect. The fourth had a weird printing flaw.
The AI chat agent located my order and confirmed the mistake in under two minutes. After a quick photo upload, they sent me a replacement — no return required.
No friction. No unnecessary steps. And faster than speaking with a human. Win, win. I’ve ordered from the site since and will continue to.
When AI Adds Friction For the Customer
A few days later, I went to 1-800-Flowers.com to report that a centerpiece arrangement I’d ordered for my mother-in-law showed up looking… shall we say, lackluster. It was more of a Charlie Brown Christmas tree than the real thing!
Here’s the worst part. There was a Home Depot tag still attached to what was supposed to be a florist-designed arrangement.
Here’s the side-by-side of what I ordered in the app and what my mother-in-law received.
The chatbot experience on their site was slow and painful. For one thing, there was no way to upload a photo. The clunky, outdated AI-powered decision tree eventually routed me to a live agent.
That agent was so bad that I asked in the chat multiple times, “Are you AI?” The second time she confirmed she was a human agent (after saying, “That’s great news!” when I told her my mother-in-law had found a Home Depot price tag on the arrangement), I asked her if she could transfer me to an AI agent. 😅 Yes, I would have preferred AI in this customer service interaction.
The issue was eventually resolved (after the agent tried several times to get me to accept a discount and an after-Christmas redelivery). But, it was slow, frustrating, and annoying. Especially after I’d had such a quick, seamless experience just a few days before with another chatbot.
When AI Knows When to Involve A Human
An example of AI in customer service done intentionally comes courtesy of Minted.com. I’ve used Minted for years to order booklet-style holiday cards. This year, they accidentally mailed me the wrong cards. I received birth announcements from a family on Long Island instead of my New Year’s greetings. 🙃
What stood out wasn’t just that Minted offers 24/7 phone support. It was that their AI-powered system got me to a human in under a minute. The agent quickly apologized, refunded my entire order (completely unnecessary but deeply appreciated), and expedited a reprint with overnight shipping.
The magic of the call wasn’t that I reached a human. It was that AI knew when to get out of the way because of the issue I was calling about.
A few minutes later, I received an apology email from Minted. It talked about how important customer privacy is to them, and what steps they’d be taking to ensure it didn’t happen again.
While privacy was on my mind (I have no idea who received 150 of my family’s holiday card by mistake!), I hadn’t brought it up on the call. But, instead of shying away from that issue, they addressed it head-on. In doing so, they turned me into an even more loyal customer.
The Takeaway
From buying a TV to fixing a delivery mistake, the pattern is clear. AI in customer service works when it removes friction, speeds up resolution, and respects a customer’s time. When it introduces extra steps, dead ends, or unnecessary handoffs, it doesn’t feel innovative — it feels inconsiderate.
The technology will continue to evolve, but the standard customers use to judge their experience won’t. Customers will measure every new tool in the same way. Did this make my life easier, or did it make me work harder to get what I needed?
Enjoyed this article? I share more stories, CX lessons, and practical ideas in my newsletter. Subscribe now in the footer of my website, or go here.
Paragraph