gentleRAG

The RAG control plane for your SharePoint content

Turn enterprise documents into an AI-ready corpus that is searchable, sourced and operationally controlled — without ever losing control.

gentleRAG is an enterprise RAG suite that connects AI to the documents your teams actually use: product sheets, supplier catalogues, white papers, internal FAQs, procedures, meeting notes, in-house documentation and business knowledge bases.

A general-purpose AI can produce fluent answers. But to create real enterprise value, it must rely on your sources: your offers, suppliers, procedures, past decisions, business constraints and reference documents.

gentleRAG makes this corpus usable, observable and controlled — you decide what the AI sees, never the other way around.

Request a demo * Discuss your project

From impressive AI to useful AI

The point is not only to use ChatGPT, Claude or another model. The point is to connect AI to the living documentary knowledge of the company.

Without reliable access to internal documents, an AI answers from general knowledge. Its answers may sound convincing, while still being incomplete, approximate or disconnected from the real context.

With AI contextualized by your documents, answers become:

gentleRAG is designed to build this AI-ready reference base.

Why use a RAG approach?

An AI cannot reread all of SharePoint for every question. Volumes are too large, formats are too varied and access rights must remain controlled.

A RAG (Retrieval-Augmented Generation) approach prepares documents so AI can retrieve the relevant passages when answering.

The principle is simple:

  1. identify useful document sources
  2. prepare file contents
  3. split documents into usable passages
  4. organize these passages so they are easy to retrieve
  5. retrieve the relevant extracts for each question
  6. provide these extracts to the AI with their sources

The AI no longer answers only from its general knowledge. It answers with documentary context retrieved from your own sources.

A core principle: you stay in control

gentleRAG is not a black box that decides on its own.

It is a control room: gentleRAG — and therefore your teams — determines which files are eligible, what gets indexed, what is ignored, what should be retried or excluded. The automatic indexing and search never decide on their own that a document is valid, visible or usable.

The goal is not for the system to act in your place. The goal is to give you a clear view so you can validate, correct or relaunch at the right time.

What gentleRAG does

Control SharePoint folders

At the heart of gentleRAG is the tracked folder: a synchronized SharePoint folder used as the operational unit for document governance.

For each folder, you can see detected files, RAG-eligible documents, indexing coverage, automatic synchronization status, the latest changes, SharePoint errors, eligibility filters and the next synchronization.

Prepare documents for AI

gentleRAG handles document preparation: reading the content, text recognition on scanned files when needed, splitting into passages and making them available for search.

This step is essential. A text PDF, a scan, a sales presentation, an FAQ or a product sheet cannot be processed in exactly the same way.

Splitting is not just mechanical slicing: an FAQ must remain readable by question and answer, a procedure by step, a catalogue by product, and technical documentation by useful section.

Supervise RAG indexing

The progress of document preparation is tracked continuously: files sent, processed, successful or failed, processing in progress, estimated time and diagnostic details when something goes wrong.

Processing can be retried, cancelled or reset while keeping a clear view of history.

Qualify the corpus with AI

Once documents are indexed, gentleRAG goes beyond simple search: it identifies each document with AI to make the corpus navigable and filterable.

For each file, gentleRAG captures in particular:

This identification relies on controlled categories rather than free text, and honestly flags what it could not determine instead of inventing it.

The result: you can explore the corpus, filter by type, theme or issuer, spot related content and identify likely duplicates or competing versions.

Spot related documents and duplicates

gentleRAG compares documents based on this AI identification, not on mere text resemblance.

Two versions of the same catalogue, two variants of a procedure or two competing product sheets are brought together in a relevant way, with a proximity score — so you can clean up the corpus before widening AI usage.

Expose sourced search

gentleRAG provides a search that your AI assistants can query over the indexed corpus.

For each question, candidate passages are retrieved and then reordered from most to least relevant before being returned. Search remains limited to the selected folder and returns sourced passages, document links and a reliable document date.

Plug gentleRAG into your AI assistants

gentleRAG is not tied to any AI vendor. It exposes its corpus through MCP (Model Context Protocol), an open interoperability standard adopted by the leading assistants on the market.

In practice, you query your documents directly from the tools your teams already use — Claude (Anthropic), ChatGPT (OpenAI) or any other compatible assistant — with no custom development for each one.

A suite built from three components

The control plane

gentleRAG provides business-level control and remains the source of truth: SharePoint connections, folders, files, statuses, processing, errors, retries, resets, AI identification, related documents and usage costs.

The indexing plane

Document preparation: reading the content, text recognition, splitting into passages, making them available for search and tracking progress.

The retrieval plane

The retrieval layer lets assistants find and fetch the most relevant passages from the gentleRAG corpus, with identified sources and an explicit document scope.

Hosting: à la carte, to fit your constraints

gentleRAG is modular. Each component can be hosted wherever it makes the most sense for you, to balance sovereignty, confidentiality, performance and cost:

You choose the hosting model, component by component:

These options combine: you could, for example, keep the document database and indexing in-house for confidentiality while entrusting the control plane to gentleStacks for simplicity.

Tangible benefits

1. AI grounded in your documents

Answers no longer rely only on the model's general knowledge, but on your procedures, product sheets, catalogues, FAQs, meeting notes and business documents.

2. Faster access to information

Teams can quickly access consolidated information across several documents or silos: sales, support, quality, suppliers, projects and management.

3. A stronger company memory

Past decisions, meeting notes, lessons learned and internal documentation become easier to reuse in new analyses.

4. Faster onboarding

New team members can find the right documents, reference procedures and useful history without depending only on internal experts.

5. Clearer document operations

Teams know exactly what is synchronized, indexed, ignored, in error or waiting for a retry — and an AI-identified corpus can be browsed by type, theme or issuer.

6. Traceable answers

Sources, document links and dates make it possible to verify where an answer came from and update the source document when needed.

7. Controlled, visible AI costs

Each AI usage is measured separately — document identification, corpus preparation, search — to keep a clear view of spending and avoid costs that quietly spiral.

Governance and document freshness

An AI document base should not be a one-off import of files.

gentleRAG is designed to track source evolution: additions, changes, deletions, reindexing, errors, manual exclusions and status changes.

Index freshness directly affects answer reliability. A deleted, replaced or obsolete document must not keep feeding the AI as if it were still valid. When a document changes, its AI identification and its matches are refreshed so they never stay on an outdated analysis.

Manually excluding a file keeps the document in SharePoint but cleanly removes it from AI processing until it is reincluded.

The goal is not for the system to decide on its own. The goal is to give teams a clear view so they can validate, correct or relaunch processing at the right time.

Typical use case

A company has thousands of files in SharePoint: supplier product sheets, catalogues, white papers, internal FAQs, procedures, contracts, meeting notes and training material. With gentleRAG, it selects the folders to track, synchronizes their content, indexes useful documents, monitors errors, identifies files with AI and exposes sourced search to its assistants.

Result:

In summary

gentleRAG is for organizations that want to move from general-purpose AI to genuinely useful enterprise AI.

The suite helps you:

Contact

Let us talk about your document corpus

Would you like to turn your SharePoint folders into an AI-ready knowledge base?

gentleStacks helps you scope, prototype and deploy a RAG architecture adapted to your content, use cases and constraints.

Contact: Get in touch