RIP Customer Service

1. What happened

Customer service used to be a relationship.

You had a problem. You called. A person answered. They fixed it or they didn't. Either way, you spoke to a human who could make a decision.

That model is dead.

It was not killed by technology.
It was killed by incentives.

2. The new model

Many large companies now route customer complaints through a system designed to do one thing:

Make resolution more expensive than giving up.

The tools are familiar:

• Chatbots that cannot resolve anything
• Phone trees that loop
• Hold times measured in hours
• Agents with no authority
• Ticket systems that lose context
• Escalation paths that do not escalate

Not every company does this. Some get it right. But enough do it that the pattern is unmistakable.

3. The economics

Customer service is a cost centre.

Every contact costs money. Every resolution costs more.

The rational move is:

• Reduce contacts
• Deflect complaints
• Delay resolution
• Exhaust the customer

Most customers will give up before the company's support system gives in.

This is not a failure.
It is triage by frustration tolerance.

4. Why it got worse, not better

Technology made customer service cheaper to deliver badly — but consistently.

• Chatbots cost less than call centres
• Automated responses cost less than chatbots
• No response costs nothing

Of course, the cheapest customer support is not creating the need for support in the first place. But that requires caring about the product more than the margin.

Each generation of tooling reduced the cost of deflection while maintaining the appearance of availability.

The company still has a "Contact Us" page.
It may or may not lead somewhere useful.

5. The complaint rate trick

Companies measure complaint volumes.

When complaint volumes drop, they report success.

But volumes can drop for two reasons:

• The problem was solved
• Customers stopped trying

Most complaint reduction is the second kind.

Activity is measured. Outcome is not.
The metric improves while the problem worsens.

6. The human cost

Customers are not just inconvenienced.

They are conditioned.

After enough failed attempts to resolve a problem, most people learn:

• Calling is pointless
• Complaining is exhausting
• Switching feels risky
• Acceptance is easier

This is learned helplessness applied commercially.

The company did not solve the problem.
It solved the complaining.

At one point, the Commonwealth Bank's call wait time to reach a human was approximately eight hours. That meant you could not get through before the bank closed, so you were forced to use CEBA — their AI chatbot. CEBA is about as smart as a bag of rocks and has no power to resolve anything. The loop was complete: you could not reach a person, and the machine could not help.

7. Why frontline staff cannot help

The people answering phones are not the problem.

They are:

• Underpaid
• Undertrained
• Given scripts instead of authority
• Measured on call time, not resolution
• Penalised for escalating

The system is designed so that the person you reach cannot help you, and the person who could help you cannot be reached.

But this is not universal.

Amazon's customer service is generally good — they resolve most issues quickly, with real authority at the front line. The exception is IT-related problems, where they fall into the same loops as everyone else.

AliExpress is the opposite. Support is slow, formulaic, and often useless.

eBay has taken a different path entirely: they offer almost no customer support at all, except to refund you if you cannot work it out with the vendor directly. It is a system that outsources resolution to the marketplace.

The variation matters. It proves this is a design choice, not an inevitability.

8. The Fuckwit layer

Between the customer and anyone with authority sits a layer of middle management optimising for:

• Cost per contact
• Average handle time
• Deflection rate
• Complaint closure rate

These metrics reward exactly the wrong behaviour.

A complaint closed without resolution counts the same as one closed with resolution. A customer who gives up counts the same as one who is satisfied.

The Fuckwit does not need to lie.
They just need to report the metric.

9. What AI changes for customers

We predict the rise of the personal AI complaint and support bot.

Your AI agent will not:

• Get tired
• Feel awkward
• Accept the first answer
• Lose track of a case
• Give up after forty minutes on hold

It will:

• Document every interaction
• Escalate systematically
• File regulatory complaints automatically
• Switch providers without hesitation
• Publish outcomes at scale

The company's entire deflection architecture is built to exhaust humans.

AI does not exhaust.

10. The asymmetry flips

Right now, the company uses automation against the customer.

AI gives the customer the same weapon.

• Company deploys a chatbot → customer's AI navigates it instantly
• Company creates a phone tree → customer's AI maps it and bypasses it
• Company delays response → customer's AI follows up continuously
• Company offers a bad resolution → customer's AI benchmarks it against the market and rejects it

For the first time, the cost of deflection rises faster than the cost of resolution.

11. What companies will be forced to do

When AI agents handle complaints:

• Deflection stops working
• Bad resolutions get published
• Churn becomes instant
• Reputation damage becomes measurable and attributable

Companies will face a choice:

• Solve problems quickly and cheaply
• Or pay more to not solve them than it would have cost to fix them

The economics flip.
Resolution becomes the cheaper option.

12. Why this is Paragentic

Customer service died because passivity was profitable.

Paragentism does not ask companies to be nicer.

It changes the conditions:

• Complaints that cannot be ignored
• Exits that cannot be prevented
• Outcomes that cannot be hidden

No regulation required.
No outrage required.
Just agency, delegated and persistent.

13. Final line

Customer service was not murdered by greed.
It was optimised away by a system that discovered most people will accept being ignored.

AI does not accept being ignored.

That changes everything.