Anonymize CRM data for AI
The need to anonymize CRM data for AI concerns every sales, sales administration or customer service team that wants to summarize an exchange, write a follow-up or analyze a CRM export. Quotes, orders, invoices and customer records contain contact details, negotiated terms, discounts and volumes, information that falls under the GDPR for individuals and under trade secrets for the company. Biffer processes these files on the salesperson's computer, with no connection, and replaces each identifier with a stable alias.
This page describes the files concerned, the applicable rules, the AI uses and the fields of the sales profile.
Quotes, orders, invoices, CRM exports: typical sales files
The sales cycle leaves a paper trail at every stage, and every trace is named.
- Quotes and sales proposals (Word, PDF): quote number, contact person, discounts granted, special terms.
- Orders and delivery notes: order number, delivery number, delivery address, name of the recipient.
- Invoices and credit notes (PDF): invoice number, customer number, French company number (SIRET), IBAN for payment, due dates.
- CRM exports (CSV, Excel): contacts, emails, phone numbers, opportunity number, pipeline amount, negotiation stage.
- Meeting reports and email exchanges: needs expressed, objections, competitors mentioned.
Biffer processes PDF, Word, Excel, CSV and text files while preserving the formatting: a CRM export keeps its columns, a quote keeps its tables. A purchase order scanned as an image goes through built-in optical character recognition, offline, then is redacted by blacking out the pixels, with no possible restoration. The general method is described on the page prepare a document for AI.
GDPR, prospecting and trade secrets: sales obligations
The contacts in a customer file are individuals, even in B2B: a buyer's name and work email are personal data. In France, the CNIL (the French data protection authority, page in French) regulates their use in prospecting and reminds companies that they must secure this data and entrust its processing only to providers bound by contract, which applies to an online AI service.
On the company side, negotiated terms, discounts and volumes are trade secrets (French Law No. 2018-670 of 30 July 2018), provided they are subject to reasonable protection measures. Confidentiality clauses signed with key accounts often go further. Three consequences.
- A customer or order number leads straight to the record in the ERP: it identifies the customer company and, by extension, its contacts.
- Amounts and discounts, cross-referenced with the customer's name, reveal your pricing policy; pseudonymizing the customer is often enough.
- Where processing happens matters. With offline software, files stay inside the company and the processor question is settled in advance.
Biffer prepares, you validate: the salesperson reviews the preview and the company has the final word on what it shares.
AI use cases and the Biffer workflow
Once customers and contacts are replaced, the AI becomes a good sales assistant, because writing and analysis concern the substance of the exchange rather than the customer's identity.
- Write a follow-up or a reply to an objection from a meeting report.
- Summarize a three-month email thread with a key account.
- Analyze a pipeline export in CSV to spot stalled opportunities.
- Rework a quote into a structured sales proposal.
The workflow is the same every time.
- You drop the file into Biffer with the sales profile enabled.
- The local analysis highlights people, emails, phone numbers, addresses, SIREN and SIRET, IBANs, customer, quote, order, invoice, opportunity and delivery numbers.
- You untick false positives (a product name, an item reference) and add what is missing, all occurrences at once.
- The document comes out with stable aliases (CLIENT_001, DEVIS_001 for a quote, PERSONNE_003; the alias labels are in French); in Word and PDF, redacted data is highlighted in black, an option enabled by default.
- The mapping table, encrypted or in plain text, restores the real names in the follow-up written by the AI, ready to send.
The queue processes a batch of files with the same aliases from one to the next. Biffer is free for five files per month, after a one-time email confirmation in the software; beyond that, Biffer for Mac or Biffer for PC costs €10 excl. VAT (€12 incl. VAT) per computer, paid once: lifetime license, unlimited files, all redaction features included (see the pricing page). The steps are detailed on the features page.
The fields the sales profile recognizes
The sales profile adds the numbers of the sales cycle to the general detections. Here is the exact list, as it appears in the software.
| Field | Where you find it |
|---|---|
| Customer number | Invoices, CRM records, letter headers |
| Quote number | Quotes, proposals, purchase orders that refer to them |
| Order number | Order acknowledgments, delivery notes, invoices |
| Invoice number | Invoices, credit notes, reminders, accounting exports |
| Opportunity number | CRM exports, pipeline reviews, meeting reports |
| Delivery number | Delivery notes, carrier tracking, disputes |
On top of these come the four fields shared by all professions (file number, reference, contract number, employee ID) and the general detections: people, emails, phone numbers, postal addresses, SIREN and SIRET, French social security number, IBAN, license plates, dates. For an identifier specific to your ERP, for example "Affaire AFF-2026-0215", the custom keyword field described on the professions page does the rest. The accountant profile can be added if needed. Biffer is published by Romain Rissoan, trainer, consultant and founder of OPCADIA, in Lyon.
Frequently asked questions
Are amounts and discounts redacted?
They stay in place by default. Biffer handles the data that identifies a person or a company: names, contact details, SIREN, customer, quote or order numbers. An amount or a discount percentage identifies no one as long as the customer is replaced by its alias, and the AI needs these figures to analyze a pipeline or rework a proposal. If a price list should stay inside the company, you can create a custom keyword field ("Discount:", "Net price:") so that the values that follow are replaced, or simply delete the column before processing.
How do you handle a CRM export with several thousand rows?
A CSV or Excel export is processed cell by cell, on your computer. Each contact gets a stable alias (PERSONNE_001, EMAIL_001) that stays identical at every occurrence, including in other files processed in the same queue: a meeting report that mentions the same buyer will get the same alias as the CRM row. We advise validating the detections on an excerpt of a few dozen rows in reviewed mode, then running the full file. Columns, row order and Excel formulas are preserved, which lets the AI calculate totals by stage or by salesperson without knowing the customers.
Can the follow-up written by the AI be sent back with the real names?
Yes, that is the purpose of the mapping table. At export, Biffer offers to save it, encrypted (Argon2id and AES-256-GCM) or in plain text depending on how well protected your computer is. When the AI returns a follow-up addressed to PERSONNE_001 about DEVIS_002, you load that text back into Biffer with the table: the real contact name and quote number are restored, locally, and the letter is ready to go. Without a table, the pseudonymization is irreversible, which suits a pipeline analysis whose result never needs to go back to customers. Either way, everything stays on your computer.
Your business documents anonymized, ready to be published or handed to an AI tool.
Biffer runs on your computer without sending a single file. Free for five files a month, unlimited for €10 excl. VAT per computer.
100% offline · Free: 5 files a month · No file sent