In 2008, musician Dave Carroll watched through the airplane window as United Airlines baggage handlers threw his $3,500 Taylor guitar onto the tarmac. When he retrieved it, the neck was broken.

He spent nine months trying to get United to pay for the repair. The claims process bounced him between departments. Nobody took responsibility. He asked for $1,200 in flight vouchers to cover the repair cost. United refused, saying he hadn't filed a claim within 24 hours.

So he wrote a song. "United Breaks Guitars" went up on YouTube in July 2009 and got 3 million views in its first week. It has over 25 million views as of 2026.

Within weeks of the video's release, United's stock price dropped 10%, wiping out approximately $180 million in shareholder value. The direct causation is debated (United's stock was already struggling), but the PR damage was undeniable and the $180 million figure became a cautionary benchmark.

All over a $3,500 guitar repair.

The Comcast Retention Call

In 2014, tech entrepreneur and former Engadget editor-in-chief Ryan Block tried to cancel his Comcast service. The retention agent refused to process the cancellation, repeatedly asking why he wanted to leave and arguing with his answers. The call lasted about 18 minutes total.

Block recorded the last 8 minutes of the call and posted it online. It went viral. Comcast was already one of America's most disliked companies, and this recording became the definitive proof of everything people suspected about their cancellation process.

The FCC received a spike in complaints. Comcast's already-low customer satisfaction scores dropped further. The company eventually apologized, but the damage was done. "Comcast retention call" is still referenced in customer service training a decade later.

The agent was following the script. The script was the problem.

What These Stories Have in Common

The guitar cost $3,500 to fix. The Comcast cancellation would have cost $0 (the customer was leaving anyway). In both cases, the company chose to fight over a tiny amount and lost millions in brand damage.

The pattern repeats across industries:

A bank refuses to waive a $35 overdraft fee for a customer with $200,000 in deposits. The customer moves everything to a competitor.

An airline won't rebook a passenger whose connection was delayed by the airline's own late departure. The passenger live-tweets the experience to 50,000 followers.

A SaaS company takes 5 days to respond to a trial user's question. The user writes a detailed "why I didn't buy [product]" post on Hacker News. It hits the front page.

The common thread: the cost of doing the right thing was trivially small. The cost of doing the wrong thing was enormous. And nobody did the math beforehand.

The Asymmetry of Bad Service

Good support rarely goes viral. A customer who gets a fast, helpful response might tell one or two people. A customer who gets terrible support tells everyone.

Research from the White House Office of Consumer Affairs (now dated but consistent with newer studies) found that unhappy customers tell 9 to 15 people about their experience. With social media, that number is functionally unlimited.

One bad experience doesn't just lose one customer. It becomes content. A tweet, a TikTok, a Reddit post, a Hacker News thread. Each one reaches hundreds or thousands of potential customers who now associate your brand with that experience.

The flip side: one great recovery can also go viral, but only if it's genuinely remarkable. Ritz-Carlton's $2,000 authorization for any employee to solve a guest problem without approval generates positive stories regularly. But "Ritz-Carlton did something nice" is news because it's unusual. "Company was terrible" is news because people expect it and love confirming their expectations.

The Math Companies Should Do

Before denying a claim, refusing a refund, or making a customer jump through hoops, do this calculation:

Cost of just fixing it: usually $10 to $500.

Cost of the customer churning: their annual value to you, typically $100 to $10,000.

Cost of a negative public review: $500 to $3,000 in lost future revenue (Moz and industry estimates).

Cost of a viral incident: incalculable. Could be nothing. Could be $180 million.

Expected value of denying the claim: (probability of no blowback × $0 saved) + (probability of churn × churn cost) + (probability of public complaint × review cost) + (probability of viral × massive cost).

The expected value of denial is almost always negative. The expected value of a fast, generous resolution is almost always positive.

How AI Prevents Disasters

Most support disasters don't start as disasters. They start as routine interactions that go wrong because of slow responses, bounced tickets, rigid policies, and agents who don't have authority to fix simple problems.

AI prevents disasters by eliminating the early failures:

Speed. The customer gets a response in seconds, not days. The frustration doesn't build. Carroll spent nine months trying to reach someone at United. If he'd gotten a real response in day one, there would be no song.

Routing. The message goes to the right person immediately. No bouncing between departments, no "that's not my department," no transfers that lose context.

Authority. For issues under a certain dollar threshold, AI can resolve them automatically. Refund the $35 fee. Send the replacement item. Process the credit. No escalation, no approval chain, no delay.

Detection. AI classification can identify high-risk messages (threats to go public, mentions of lawyers, references to social media) and route them to senior staff immediately. The Comcast scenario would have been caught by an AI that flagged "I've been trying to cancel for 8 minutes" as an escalation trigger.

The irony of United's guitar story is that the cost of fixing it was $3,500, and the damage was $180 million. That's a 51,000x multiplier. Every company sitting on unresolved support tickets has some version of this ticking time bomb in their queue. You just don't know which one it is.

The safest bet: fix them all fast.