Customer Experience

I Was Ready to Pay. The Bot Would Not Let Me.

Twelve days, three channels, one promised follow-up that never came. What a failed purchase taught me about customer service in the AI age, and why the fix is not less automation but better design.

The purchase that never happened

At the end of October I set out to buy several home appliances from a well-known retailer. The items were already sitting in the cart on their own website, and my card was ready. One order, paid in full, no negotiation needed.

I was not asking for credit, a favor, or a special deal. I wanted to pay them. Here is how that went.

Days 1 to 4 - WhatsApp

The automated assistant asked for my name, ID number, items, and delivery address. Then it asked again. It never answered the four questions I actually had: product details, warranty, return procedure, and which human I could talk to.

Same week - Phone

I called. Nobody picked up.

November 2 - Email

I sent one complete message: a screenshot of the cart, the item list, the delivery address, two contact people, my four questions, and a request for payment options and any volume discount.

November 3 - Follow-up

"This is the email I have not received an answer to." Silence.

November 9 - WhatsApp, one last try

I sent the same information again. This time they confirmed they had received it, and told me customer service would follow up to confirm delivery and the payment method.

November 10 - Last message

After giving them a full day to get back to me, I sent a cordial email letting them know I would move forward with another provider, and offered to help them improve their process. As I write this, still no reply.

Three channels. A customer with a full cart and a card in hand. One promise to follow up, and no follow-up.

The bot was not the problem

I build AI assistants for a living, so I am not going to tell you chatbots are the villain here. The bot did the easy part well: it collected my data. Then it collected it again.

What it could not do was the only thing that mattered. It could not hand me to a person who could take my money. And when I went around it, every other channel failed the same way: nobody owned the outcome.

That is the real failure pattern I see in small and mid-size businesses rolling out AI customer service. The company buys a bot to answer faster, and quietly treats it as a reason to stop watching the other doors. The bot becomes the wall, and the inbox becomes a place where messages go to die.

What customers have been telling us

None of this surprises researchers. In a Gartner survey of 5,728 customers, 64% said they would prefer companies did not use AI for customer service, and 53% said they would consider switching to a competitor if a company was going to use it. The top concern, cited by 60%, was that AI would make it harder to reach a person.

That survey was fielded in December 2023, so it is not new. My experience says the fear is still valid.

But here is the other side. Zendesk's CX Trends 2026 report found that 77% of customers find chatbots helpful for simple issues, and 71% agree AI can deliver faster replies. Customers are not rejecting automation. They are rejecting being trapped by it.

Customers are not against bots. They are against bots with no exit.

The companies that went furthest with "replace the agents" are already walking it back. In June 2025, Gartner predicted that by 2027, half of the organizations that expected to significantly reduce their customer service workforce will abandon those plans. In a poll of 163 service leaders, 95% said they plan to keep human agents.

Five rules we build into every customer-facing assistant

At Ideas Factory, we design the handoff first and the bot second. These are the rules I would have wanted that retailer to follow, and the ones we hold our own builds to.

1. An exit to a human within two failed turns

   If the assistant cannot answer twice in a row, it offers a person, with a name, a channel, and a realistic time. Not "someone will contact you."

2. Never ask twice

   Everything the customer already gave travels with the conversation into the handoff, the ticket, and the CRM. Asking me for my ID a second time told me nobody was reading the first answer.

3. Every channel has an owner and a clock

   If you publish a WhatsApp number, a phone line, and an email address, each one needs a named owner and a response target. An unmonitored inbox looks exactly like a company that does not care.

4. Let the bot close, not just collect

   When a customer says "I want to buy this," the assistant's job is to move the sale forward: a quote, a payment link, or a salesperson on the line. A bot that gathers data from a ready buyer and then stalls is a lead-loss machine.

5. Close the loop every week

   Every failed conversation is free research. We review handoffs, unanswered questions, and abandoned purchases weekly, then feed them back into the assistant's knowledge and into the business process. That feedback loop is what makes the experience better over time instead of just faster.

What this actually costs

For that retailer, the direct cost is one complete order from a customer who had already decided to buy. That is the small number.

The bigger numbers are harder to see. The two people I copied on my last email now know the story. So does everyone I talk to about where to buy appliances. And every other customer who hits the same dead end disappears without showing up in a report, because a sale that never starts leaves no data.

That last point matters for any small business. Your dashboard shows the conversations your bot completed. It does not show the customers who gave up.

What good looks like

A good AI customer experience is not complicated. The assistant answers the simple questions instantly, around the clock. It recognizes buying intent and moves it forward: payment link, quote, or a person. It hands off with full context when it is out of its depth. And someone on the business side reads the misses every week and fixes the cause, not just the script.

That is the standard we hold ourselves to at Ideas Factory, from the assistants we build to the processes behind them. The technology is the easy part. Ownership is the hard part, and it is the part your customers feel.

I ended my last email to that retailer with an offer to help. I meant it. The fix is not expensive. It just has to be designed on purpose.