Designing AI for When Things Go Wrong

What a travel fiasco taught me about AI, empathy, and designing for real-world complexity.

After three wonderful weeks in Europe, my family and I were ready to fly home.

Instead, we found ourselves trapped in a glass jet bridge in 100+° heat, waiting for a British Airways flight that never came. No updates. No clear answers. Just hours of confusion, missed connections, meltdowns — and eventually, a full rebooking out-of-pocket after we gave up and found our own way home.

I understand delays happen. But this was something else: Multiple cancellations, repeated delays, no communication, and dwindling alternatives. It was a system that couldn’t see our situation, connect the dots, or respond with any kind of coherence.

This is precisely where AI could make a real difference in customer service. Not just as a chatbot or cost-saving measure, but as a system-wide problem solver. A tool that sees the whole picture, understands real-world context, and helps both agents and customers navigate complexity when they need it most.

Where AI Could Have Helped

In our case, every support channel failed. The gate agents wouldn’t (or couldn’t) help us rebook. Phone support was closed. The website and chatbot were useless. There was no one, no tool, capable of helping us figure out what to do next.

British Airways’ entire support infrastructure was built around rigid workflows and narrow rules. There was no system-level awareness, just isolated scripts that broke the moment things got complicated. Nothing was able to zoom out, understand our context, and offer a viable solution.

Yet this is precisely where AI could be invaluable. Imagine a system that could:

• Understand a customer’s full itinerary, even across separate bookings

• Factor in urgency, delay patterns, and possible alternatives

• Explore paths creatively, rather than just following predefined logic

• Help agents communicate options clearly and empathetically

• Escalate or rebook proactively, rather than reacting too late

Imagine AI not just for automation, but for systems thinking and problem solving.

Toward a More Thoughtful Use of AI

To be fair, AI already plays a role in customer service today and shows up in many places, including:

1. Front-end tools – chatbots, phone trees, and help flows

2. Agent assistance – copilots that suggest replies or surface documentation

3. Back-end automation – ticket tagging, routing, sentiment scoring

4. Utility tools – translation, summarization, dynamic personalization

While these tools do improve certain tasks, they are designed for efficiency, not empathy. It’s not better automation that excites me; It’s the potential to bring more context, flexibility, and responsiveness into systems that desperately need it.

As a customer, it often feels like you’re up against a system that doesn’t understand your needs and isn’t trying to. But AI could:

• Support agents by making complex policies easier to navigate

• Advocate for the customer by understanding intent

• Help both sides get to a resolution faster (and with less emotional wear-and-tear)

• Make it easier for exceptions to be made when they’re clearly the right thing to do

To get there, we have to rethink what we’re optimizing for. Right now, most systems chase speed and ticket deflection. What if we instead designed for empathy, clarity, and resolution?

What It All Comes Down To

We need AI systems that are connected—systems that understand the customer’s context, see the full journey, and reason across company policies, constraints, and priorities.

AI that isn’t just fast, but thoughtful.

That doesn’t just reference documentation, but solves problems.

That bridges the gap between what the customer needs and what the business can offer.

The future of AI and customer service isn’t about replacing people. It’s about designing tools that can see the whole system — and help everyone (including agents and customers) reach the best outcomes.