Ryan Block wanted to cancel his Comcast account. Simple enough. He called the number, navigated the phone tree, and reached a retention specialist.

The call lasted about 18 minutes. About 10 minutes in, Block started recording. In the 8 minutes of audio he captured, the agent repeatedly refused to process the cancellation. He asked why Block wanted to leave. Block said he just wanted to cancel. The agent pushed back: "Why don't you want the fastest internet in the country?" Block repeated his request. The agent argued. Block stayed calm. The agent got more aggressive.

The recording went viral. Millions of people listened. Comcast apologized and said they would "take quick action." Block himself publicly asked Comcast not to fire the agent, saying the problem was systemic, not individual.

But here's what everyone missed: the agent was following the script. Comcast's retention training explicitly instructed agents to overcome objections and save the account. The agent did exactly what they were trained to do. The training was the problem.

Why Retention Scripts Fail

The Comcast model (and variants of it used by gyms, cable companies, newspapers, and subscription services) assumes that if you just keep the customer on the phone long enough and offer enough discounts, a percentage will stay.

And it's true. Aggressive retention saves 10 to 20% of customers who call to cancel. The math looks great in a spreadsheet.

What the spreadsheet doesn't capture: the 80% who still cancel are now furious. They tell everyone. They leave negative reviews. They become anti-advocates. The brand damage from aggressive retention far exceeds the revenue saved from the 10 to 20% who stayed.

Block's recording didn't just embarrass one agent. It confirmed what millions of people suspected: that cancelling a service is intentionally made difficult. Every company that uses aggressive retention tactics gets painted with the same brush.

The FTC Noticed

In 2024, the FTC finalized its "click to cancel" rule. If a customer can sign up online, they must be able to cancel online with equal ease. No mandatory phone calls. No retention gauntlets. No "let me transfer you to a specialist who will try to talk you out of it."

This rule exists because of experiences like the Comcast call. Millions of them.

Violations carry fines of $50,000+ per incident. More importantly, the rule reflects consumer expectations that have shifted permanently. Customers expect easy cancellation. Companies that still make it difficult are swimming against a legal and cultural current.

What to Do Instead

Make cancellation easy. A "Cancel" button in account settings. Two clicks max. No phone call required. No mandatory retention conversation.

Ask why (optionally). After the customer clicks cancel, show a short survey: "Mind telling us why?" with 4 to 5 options plus "other." Make it optional. Don't gate the cancellation behind it.

Offer one relevant alternative. Based on their stated reason, offer one thing: a pause, a downgrade, a discount. One offer, once. If they decline, process the cancellation immediately.

Be gracious. "Your subscription has been cancelled. You'll have access until [end of billing period]. We'd love to have you back anytime." That's the entire exit experience.

The customer who cancels gracefully might come back. The customer who cancels after a 10-minute argument will never come back and will actively warn others away.

AI Gets This Right

AI has no ego. It doesn't take cancellation personally. It doesn't have a retention quota to hit. It follows the process: ask the reason, offer one alternative, process if declined.

Supp classifies "cancel" intents and can distinguish between hard intent ("cancel my account right now") and soft intent ("I'm thinking about cancelling"). Hard intent goes straight to processing. Soft intent gets routed to a human who can have a genuine conversation about what's wrong.

The distinction matters. A customer who says "cancel my account" has decided. Respect that. A customer who says "I'm not sure this is worth it anymore" is open to discussion. AI knows the difference because the language patterns are distinct and classifiable.

The Comcast call happened because a human agent couldn't (or wouldn't) distinguish between these two states. The customer had decided. The agent treated it as an objection to overcome. AI doesn't make that mistake.