Why clean before you analyze
Quality teams usually review only a sample of calls. With AI, you can read all of them. But each conversation contains a customer's data, and sending everything to a model multiplies the exposure.
In collections, the conversation also covers debts and income. Under Brazil's General Data Protection Law (LGPD), the company is responsible for the data that it shares with suppliers.
To summarize, classify the reason and measure quality, the AI does not need to know who the customer is. It needs what was said.
How it works
The conversation arrives through the API
Your system sends the call audio to the transcription route, or the chat and email text to the cleaning route. A short list of texts goes in one request.
Rota removes the personal data
Names, CPF and CNPJ numbers, phone numbers, emails, addresses, and protocol and registration numbers become placeholders, such as [PESSOA_1] and [CPF_1].
AI analyzes the conversation
The model summarizes, classifies the reason, checks the script and suggests the next step. It receives only the clean text.
Your dashboard joins the result
Your system links the analysis to the contact through the identifier that it already has. The Rota dashboard shows the usage and the categories found, without the values.
Example with fictitious data
Before Rota
Cliente: Meu nome é Marcos Vinícius Teixeira, CPF 111.444.777-35. A fatura veio em dobro este mês. Podem me ligar no (11) 90000-1234 ou escrever para marcos.teixeira@example.com.After Rota
Cliente: Meu nome é [PESSOA_1], CPF [CPF_1]. A fatura veio em dobro este mês. Podem me ligar no [TELEFONE_1] ou escrever para [EMAIL_1].How to integrate
For text, use the cleaning route. It returns the clean text in clean_text and the categories found in findings. For a short list of messages, send items instead of text.
For calls, use the transcription route: the text comes back clean through the same barrier.
curl https://rota-nacional.ia.br/v1/privacy/clean \
-H "Authorization: Bearer $ROTA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "Meu nome é Marcos Vinícius Teixeira, CPF 111.444.777-35."}'Good to know
- Start with the free trial, with conversations from your process.
- For a high call volume, the team agrees the capacity with you.