Hello. Your Call Is Very Valuable to Us.

No, you heard that correctly.

We used to say “important” to avoid discussing money.

You are calling to get your nineteen ninety-nine back. Along the way, you explain what made you angry, why you no longer trust us and what we would have to do to make you stay.

Companies normally pay for customer research.

You are providing yours free. Sometimes while paying for the call.

Thank you. Please hold.

There is a particular kind of relationship in which you explain your problem, someone listens carefully, asks thoughtful questions and says they understand completely.

Then they ask you to wait.

You wait.

They return and suggest telling the whole story to someone else.

If dates worked like this, we would have stopped leaving the house. But this is customer support. Here, humiliation comes with a reference number.

Now it has an investment angle, too.

On 5 November, Five9 announced an agreement to acquire Level AI for approximately $325 million in cash. The transaction has yet to close. Level AI builds tools for analysing customer conversations, assisting agents and automating service. According to the announcement, its platform handles more than a billion customer interactions a year.^[1]

How many begin with “I already explained this to your colleague” is not disclosed.

But humanity appears to have found a common language.

Remember Mickey Mouse and our “nudes”? Bare souls hoping someone would finally answer.

That story involved twenty million dollars.^[2]

This one involves three hundred and twenty-five.

Apparently, your problems are worth more than your nudes.

It makes sense. A bare soul still needs persuading to open up. An unhappy customer will call you themselves.

They will tell you what happened. Why it hurts. What you promised and failed to deliver. What might persuade them to stay.

If you do not understand, they will repeat it.

You pay a therapist for this much honesty. Here, at least someone else can make money from it.

And there is no painful hunt for research participants.

Just fail to refund someone.

They will insist on being heard.

You can write code. Getting an honest answer from a person is harder.

Invite them to an interview: they will try to sound reasonable. Give them a questionnaire: they will choose the middle of the scale. Ask how they feel about your brand: they will suddenly remember an appointment.

Get their bill wrong: they will find the time.

Not every call is a revelation. But a meaningful conversation offers something a button marked “rate us from one to five” cannot: an explanation.

What someone considers fair. Which promise you broke. Where their patience ends.

That is why I am interested in more than the team that built the system. I am interested in the people giving it something to analyse.

They did not apply to work there.

Five9 describes turning customer conversations into insights the whole organisation can use.^[1]

This is where our plans diverged slightly.

I intended to explain to one person why my parcel had not arrived. The conversation had ambitions.

That can be useful. If a thousand people complain about the same mistake, someone ought to notice it.

But a remarkable number of presentations fit between “we identified your pain” and “we fixed it”.

My problem has reached the strategy meeting. It has not reached the refund button.

I opened the documents of companies publicly listed as Level AI customers. They are more interesting than the hold music.^[1][6]

VistaPrint describes recording calls and chats, AI suggestions for support staff, feedback analysis and interaction summaries. Communication information also appears among the data categories used for research and service development.^[3]

You hang up. The conversation goes back to work.

I thought we were discussing badly printed business cards. Apparently, the conversation had a career plan.

Gusto, the payroll and HR platform, explicitly prohibits third-party AI providers from using personal information to train, fine-tune or improve their own models. That is a meaningful restriction.

But it describes separate rules for optional AI integrations. Disconnecting stops future disclosures. Information already shared remains subject to the provider’s privacy policy and retention practices.^[4]

We broke up. His continuing relationship with my messages is governed by his privacy policy.

At OpenTable, I found a sentence worth finishing.

Third parties are not permitted to train their own models on personal or sensitive data provided to the company.

A good sentence.

You can exhale.

Except it is not finished.

An exception follows for exceptional circumstances that pass an internal review and approval process. The clause does not say whether the exception has been used, or for whom.^[5]

He promised not to tell anyone. Then we finished the sentence.

A neighbouring provision describes AI analysis of messages between diners and restaurants, with measures to remove identifying information before sharing with third-party AI services.^[5]

I only wanted a table by the window.

The terms suggested a larger party.

Processing a conversation, training a model on it and selling a recording are different acts. A public customer list does not reveal exactly which data Level AI receives. Its data-processing agreement limits processing to customer instructions, service provision and legal requirements.^[7]

But the value of a conversation is not limited to selling the file.

There is money in knowing how to find cancellation reasons, recurring complaints, effective responses and weaknesses in service.

You do not have to sell my voice to build a business around what I said.

I want my bill corrected. The company may also want to assess an employee, understand dissatisfaction and improve the product.

Sometimes our interests align beautifully.

Sometimes my complaint becomes extremely useful to everyone except me.

Then I would like to see another number beside “conversations analysed”: people who did not have to call again.

Not because they gave up. Because someone helped.

I am not against a company learning from its mistakes. Especially if it stops repeating them at my expense.

I am not against AI sparing an agent manual notes and sparing me a fifth retelling of the same story.

But if conversations have commercial value, please stop pretending the customer’s entire contribution is pressing one after the tone.

They explain, remember, clarify and correct misunderstandings. Sometimes they conduct a small investigation into why your product does not work.

Between their job, dinner and getting to school pickup on time.

You are buying a system that knows how to listen.

I am still waiting for that to help me.

Hello. Your call is very valuable to us.

Now I would like to know why I am the one paying for it.