Skip to main content

Ask Feedier and AI agents

Query your feedback in natural language, let agents monitor your data, and understand how Feedier's AI protects your data.

F
Written by François Forest

In short: AI is at the heart of Feedier, from importing a feedback to the action plan. This article covers the two features you interact with directly: Ask Feedier, which answers your questions, and AI agents, which monitor your feedback for you. It also explains how your data is protected.

AI at every step of Feedier

Step

What AI does

Learn more

Collect

Translates your surveys into every language in one click

Import

Suggests the import name, detects the date column, then analyzes each verbatim as soon as it arrives, whatever the source

Protect

Detects and masks names, places, numbers and other personal data (nlp_… attributes)

Analyze

Classifies each passage into your topics, measures sentiment, groups insights, filters out noise, and helps you build your taxonomies

Measure

Calculates sentiment and the business impact score that prioritizes your topics

Report

Summarizes and writes your reports, turns a request into filters

Act

Generates action plans with their NPS impact and effort

Query

Answers your questions and monitors your data (this article), including from Claude with the MCP server

The analysis building blocks

Feature

What it does

Where to see it

Topic detection

Classifies each verbatim into the topics of your taxonomies

Sentiment

Detects the tone (positive, neutral, negative) of each passage

Feedback feed, reports

Insights

Groups recurring findings within a topic

A topic's Insights panel

Entity detection

Spots names, places, products, brands, numbers

nlp_… attributes

Noise detection

Sets aside responses with no content (random characters, tests)

Feedback feed

Generation

Summarizes, writes reports and action plans

Reports, Action plans

Ask Feedier

The Ask Feedier panel answers your questions about your feedback in natural language: "What are the three pain points growing this month?", "What do premium customers say about refunds?".

Where

What it does

Everywhere

Answers taking into account your filters, your period and your team

Taxonomy Lab

Reviews a taxonomy, explains a classification or suggests topics. Each change is submitted to you for approval.

Filters

Turns a request into conditions ("dissatisfied customers in Chicago")

For better answers: ask a specific question, about a clear period and scope, and ask for sample verbatims.

AI agents

Agents monitor your data and notify you: an abnormal rise in a topic, a score dropping on a favorite attribute value… They suggest recommendations that you accept or reject.

Action

Who can do it

View agents and their analyses

All roles, depending on the Agents module

Turn an agent on or off

All roles, depending on the Agents module

Accept or reject a recommendation

All roles, depending on the Agents module

Anomaly detection applies to the values of attributes marked as favorites. Choose them carefully. See Attributes and referentials.

Help AI understand your business

Setting

Effect

Analyst configuration: business, goals, vocabulary

Topics, summaries and recommendations tailored to your context

Team language (Settings → Teams → Advanced)

The language of topics and suggestions

Explicit topic names and, if needed, MUST / MUST NOT rules

More accurate classification

Noise

Question

Answer

Are "noise" responses deleted?

No, they're set aside: excluded from reports, topics and scores, but still viewable.

Can I correct the AI?

Yes, any user can mark or unmark a response as noise.

Are they billed?

No.

Can I export them?

Yes, to review them. A question missing from an export may be explained by responses classified as noise.

Data protection

Feedier guarantees

Details

⚠️ Anonymization before AI

Personal data (names, emails, phone numbers, IBANs, addresses…) is detected and masked by Microsoft Presidio, running in Feedier's infrastructure in Europe. The model never receives personal data.

⚠️ No training

Providers are contractually committed not to use your data to train their models.

⚠️ Zero retention

No data is retained by the LLMs after processing, with caching disabled.

⚠️ Transparency

All AI-generated content is flagged as such in the platform.

⚠️ Regulation

Compliance with the GDPR and the AI Act.

AI providers

Provider

Mistral AI

Google

Hosting

Microsoft Azure, Sweden

Google Cloud, Belgium (europe-west1)

Models

Mistral large, medium, small

Gemini 3.5 Flash, Gemini 3.5 Flash Lite

Certifications

ISO 27001, 27701, SOC 2

ISO 27001, 27701, SOC 2

Data used for training

No

No

Depending on your plan, contact [email protected] to choose your account's provider.

FAQ

I have questions about AI security and governance

Contact [email protected] to get the compliance package.

Can AI assign the wrong topic?

Yes, just like a human. Correct a verbatim directly, or adjust the topic's rules in the Taxonomy Lab.

Does Ask Feedier use budget?

No. Only verbatim analysis is billed. See AI analysis budget.


Last reviewed: November 11, 2026, version 4.23.5.

Did this answer your question?