The CSOn platform

A complete operational intelligence platform

Native connectors, semantic knowledge base, conversational assistant and document generation — all integrated, governed and with data in Brazil.

ERP Connectors

Financial intelligence over any corporate ERP

CSOn connects directly to the Service Layer of leading market ERPs, extracting real data without changing any existing process or configuration.

SAP Business One

Accounts receivable, accounts payable, aging, tax documents and financial master data. Connection via OData Service Layer.

Available

TOTVS Protheus / Fluig

Planned integration with TOTVS financial and inventory modules, the most widely used ERP in Brazil.

In development

Oracle NetSuite / EBS

Connection with Oracle financial and operational modules, for mid-to-large companies.

In development

Other ERPs via adapter

Adapter architecture allows integrating any ERP with an API or structured export.

In development
CRM Connectors

Intelligence over the commercial pipeline and relationships

Connect your CRM to the knowledge base and CSOn assistant to cross ERP operational data with pipeline, deal history and customer relationships.

Salesforce

Pipeline, opportunities, accounts and interaction history integrated with the ERP's financial context.

In development

HubSpot

Marketing, sales and relationship data connected to the company's operational flow.

In development

Pipedrive

In-progress deal view crossed with real financial data from the ERP.

In development
Knowledge base

Semantic search across all company documents

CSOn indexes operational documents intelligently, making any information accessible via natural language — without changing where documents are stored.

Google Drive

Connect Google Drive folders directly. Documents are semantically indexed without leaving the Drive — no files are copied to CSOn's servers.

Files remain in your Drive. CSOn accesses them read-only for indexing only.

Direct upload

Upload PDFs, contracts, manuals and policies directly to the platform. Files are stored with complete isolation in the CSOn repository.

Automatic OCR for scanned PDFs. Semantic chunking and vectorization for high-precision search.

  • Semantic natural language search across the entire archive
  • Native OCR for scanned documents and image PDFs
  • Vectorization with Google text-embedding-004 embeddings
  • Results with evidence excerpt and precise location
  • Complete per-company isolation — no cross-tenant contamination
Conversational assistant

Ask in natural language. Receive answers with evidence

The CSOn assistant is not a generic chatbot. It is an auditable pipeline that semantically rewrites the question, executes queries on real client data and narrates the response with full traceability.

1

Classification

Identifies the question intent — financial, documentary, operational or conversational.

2

Semantic rewrite

Transforms the natural language question into a structured query plan.

3

Controlled execution

Queries real data from connectors and knowledge base — read-only, with restricted scope.

4

Auditable narrative

Generates the final response with source, criteria, reference date and persisted audit trail.

Document generation

Create reports and documents with AI and real data

CSOn goes beyond search. With data already indexed — from systems and documents — it's possible to generate new documents and reports using company templates, without copy-pasting between systems.

Company templates

Use the organization's existing document models as a base for automatic generation.

Real integrated data

Generated content uses real data from ERP/CRM connectors and indexed documents — not fictional data.

Controlled generative AI

Generation is done by LLM with defined scope — no hallucination of data that doesn't exist in the base.

Review and export

The generated document can be reviewed in the platform editor and exported as PDF or Word.

Ready to see the platform with your company's real data?

Schedule a 30-minute diagnosis. We'll show how CSOn fits into your current stack.