Voice of the customer: how to turn calls, interviews, and support into a knowledge base

The customer has already spoken. They spoke on the sales call, in support, in the discovery interview. They said what blocks the purchase, what almost made them cancel, the exact words they use to describe the problem. All of it was recorded. And, in most companies, it was never heard by anyone again.

The most honest voice of the customer there is — the spontaneous one, with no script, no 0-to-10 scale — lives in audio. And audio, inside companies, has a known destination: the folder of recordings nobody opens.

This post is about that hole. Why voice-of-the-customer programs collect almost everything except the actual speech. What gets lost in that discarding. And how to turn calls, interviews, and support conversations into a knowledge base that answers questions in seconds.

What voice of the customer is — and where it really lives

Voice of the customer (the famous VoC) is everything customers say about your company: needs, objections, praise, complaints, vocabulary. The term became synonymous with a formal program — satisfaction surveys, NPS, CSAT, post-support forms.

But there is a little-discussed split inside that concept:

• The solicited voice is the one you asked for. The survey you sent, the question you formulated, the scale you offered. It answers what the company wanted to ask.
• The spontaneous voice is the one the customer gave without being asked. The objection in the middle of the sales call. The long explanation to the support agent. The exact way they describe the problem — which is rarely the way marketing describes it on the website.

The spontaneous voice is richer and less filtered. And it is precisely the one that never makes it into any dashboard, because it does not live in a form: it lives in audio. Calls, meetings, interviews, voice messages. The formal voice-of-the-customer program covers the first category and ignores the second.

The hole in the middle of the program

In 2005, Bain & Company asked 362 companies a simple question: is the experience you deliver to customers superior? Eighty percent of executives said yes. When the same question reached the customers of those same companies, only 8% agreed. The “delivery gap” became a classic in the customer experience literature — and it remains current because the mechanism that produces it has not changed: the company measures what it says, not what the customer lives.

The problem is not a lack of collection. It is the type of collection. Structured research returns structured answers — clean, comparable, easy to present. And, precisely for that reason, blind to what the customer says when nobody is asking.

The industry recognizes the symptom. A KPMG study on voice-of-the-customer programs points out that 54% of customer experience leaders cannot demonstrate the ROI of their own program. Not because listening to customers does not create value — it does. The problem is that what was heard does not become a searchable asset. It becomes a report that justifies the existence of the report.

The anatomy of the discard

Let’s follow a typical sales call through the company:

1. It happens. Thirty or forty minutes of real conversation, with objections, hesitations, questions.
2. It is recorded — or not. Often not even that.
3. The salesperson writes three lines in the CRM. “Client interested, follow up in September.” The nuance died in the note.
4. The audio goes to the folder — on the phone, in the system, on the computer. There it stays.
5. Nobody replays it. And nobody ever will, because replaying costs time: a 40-minute call costs 40 minutes. A salesperson who makes ten calls a week would have to spend almost a full workday just to review their own week. A manager, to review the team’s week, would need a week. It doesn’t happen.

The result is worse than waste: it is substitution. Because the evidence is not searchable, opinion takes its place. The product meeting discusses “what customers want” based on who spoke loudest and longest. Sales decides the script by intuition. Marketing writes the page with the company’s vocabulary, not the customer’s.

The literal quote — the objection stated in the exact words, the comparison the customer themselves made with the competitor — is the most valuable raw material the company owns. And it evaporates daily, because it lives in a format nobody can search.

What AI solves — and what it should not solve

The mechanical part of this problem is solved. Transcribing a call or an interview costs almost nothing and takes seconds with AI. The bottleneck was never turning speech into text — it is what happens after the text.

And here AI helps too, but with a clear division of labor:

AI organizes. Readable transcription, faithful summary, extraction of themes and tasks, all indexed and searchable. It is volume and repetition work — perfect for a machine.

The human synthesizes. Deciding what that pattern means for the product, what contradicts the current strategy, what becomes a test. Judgment remains people’s work.

The practice of continuous discovery — the method Teresa Torres describes in Continuous Discovery Habits — points to the same place: every insight needs to stay connected to the customer and to the exact moment the speech happened. AI does the sifting and the organizing; synthesis and decision stay with the team. Automate the filing, not the understanding.

An important warning: a summary that throws away the literal quote defeats its own purpose. “The customer thought the price was high” is worth less than “you are 40% more expensive than the alternative and I can’t justify that internally.” The first is a paraphrased opinion. The second is a statement that comes back in a pricing meeting six months later with the same weight it had on the day of the call.

From audio to knowledge base: the flow

Turning calls and interviews into searchable knowledge has four moves:

1. Transcribe. The call, the interview, the support conversation becomes text with quality — correct names, numbers, and technical terms.
2. Summarize while preserving quotes. The summary points to what matters, but keeps the literal quotes anchored to the original passage. It is the difference between “the sales team’s opinion” and “evidence with an address.”
3. Organize. Themes, tasks, objections, requests — grouped and tagged. Customer A’s onboarding complaint connects to customer J’s from three weeks ago.
4. Consult. Everything saved in a knowledge base with search and chat. You ask “what did customers say about price?” and get the answers with the original quotes — not a blurred memory of someone who heard something.

This flow is exactly what Sintesy proposes: audio goes in (a call, a meeting, an interview, a voice message), and a transcription, summary, mind map, and checklist come out, with everything saved in a searchable knowledge base — in Portuguese, with search and chat on top of it all. The detail that makes the difference is not generating the summary: it is the result becoming the company’s memory, rather than one more dead conversation inside an app.

If you want the operational step-by-step of this flow for discovery interviews and calls, we already detailed it in another post: how to turn customer audio into actionable insight. This post here is about the thesis behind it: why the current discard is a strategic problem, not an operational detail.

The question test

There is a simple test to know whether your company truly listens to the customer or just says it does.

Ask: “what did twenty customers say about X?” — about price, about onboarding, about the competitor, about the reason for cancellation.

If the answer takes seconds, with quotes, with the right customer, with the exact moment of the speech — the voice of the customer has become an asset. If the answer takes a week of replaying recordings, or worse, is a gut feeling from someone who “follows this closely,” your company does not have voice of the customer.

There is voice-of-the-customer theater: a pretty survey, an updated dashboard, and the audio library — where the customer actually spoke — covered in dust.

The customer has already spoken. The only question is whether anyone will keep what they said.

Frequently asked questions

What is voice of the customer?

It is everything customers say about the company — in surveys (NPS, CSAT), on sales calls, in support, in interviews, and in voice messages. The most valuable and least used part is the spontaneous one: the speech recorded in audio, which no form captures.

Is it worth recording sales calls?

It is — it is the largest source of spontaneous voice the company has. But recording without transcribing only creates a dead archive: a 40-minute call costs 40 minutes to replay, so nobody replays it. With transcription and summary, the same call becomes a searchable record in minutes, and every objection stated by the customer stays searchable forever.

How do you transcribe a customer interview?

Record with quality, transcribe with AI, and generate a summary that preserves the literal quotes — the literal quote is the currency of customer research. Tools like Sintesy do this in a single flow: the audio goes in, and transcription, summary, and organization come out, all saved in a knowledge base with search. For analyzing meeting decisions, it is also worth reading about searchable transcripts.