Documentation

How Becide works

A practical guide to what Becide is, how it is architected, how your data is protected, and why it is different from a general-purpose AI assistant.

1Governed layer, two surfaces

What is Becide?

Becide is CFO decision intelligence for finance teams. A managed, tenant-scoped financial data lake with an AI analyst on top — it drafts your monthly management report from live curated data and answers board follow-ups in seconds, with every figure cited back to the query that produced it.

What Becide is — and isn't

  1. 01

    Fixes the data first

    GL exports, billing, payroll, spreadsheets — ingested, cleaned, and curated into one governed picture.

  2. 02

    Reads, never writes

    The product layer sits on top of governed data. It explains every number and never modifies your systems.

  3. 03

    Not a replacement ERP

    Your systems of record stay where they are. Becide reads from them, doesn't replace them.

  4. 04

    Not a DIY data lake

    Ingestion, cleaning, and curation are managed for you — no data engineering team required.

  5. 05

    Not a raw-file chatbot

    The AI only reads governed, curated layers. Never your raw files or connector credentials.

  6. 06

    Not an overseas wrapper

    Data is hosted and processed onshore in Australia — including AI inference.

Who it's for

  1. 01

    CFO

    Generates the pack in minutes and answers follow-ups with evidence during the meeting.

  2. 02

    Finance team

    One curated layer feeds both reporting and Q&A — connect once, report and ask from the same data.

  3. 03

    Board & executives

    Exported management reports plus cited answers to questions raised in the room.

  4. 04

    Administrators

    Role-based permissions, workspace scoping, and activity logs across the organisation.

Getting started

Typical onboarding is days, not months — the data foundation is provisioned as part of onboarding, not built as a project.

  1. 01

    Week 1 · Tenant provisioned

    Isolated storage, dedicated encryption key, and workspace. You connect your first data source.

  2. 02

    Week 1–2 · Data flows

    Raw, clean, and curated layers populate. A knowledge base maps business questions to your real tables.

  3. 03

    Week 2 · First pack

    Generate your first management report from live workspace data and start asking follow-up questions.

  4. 04

    Ongoing

    The pack regenerates on your reporting cadence; the assistant handles ad-hoc questions between cycles.

Product surfaces

  1. 01

    Management Report

    Board-ready monthly pack: P&L, variance to budget, segment revenue, overheads, balance sheet, cash flow — with methodology and source query links on every section.

  2. 02

    Ask Assistant

    Plain-language questions over the same governed data. Multi-step reasoning you can inspect, with citations and a methodology note on every answer.

  3. 03

    Connectors

    CSV upload and Amazon Athena in v1. Every connection is tested before saving, with schema discovery and clear error reporting.

  4. 04

    Workspace controls

    Role-based permissions, standing report instructions, connector health, and activity logs across the organisation.

Data layers

Your organisation's data lives in three layers in Sydney. AI can read only approved layers for your tenant.

RawNot exposed to AI
Exact copies of uploads. You can export or download. Never surfaced to models.
CleanAI read-only
Normalised, searchable text and vectors. Used for grounded retrieval.
CuratedAI read-only
Aggregated tables, KPIs, and reporting-ready metrics. Feeds both the pack and the assistant.

Privacy & encryption controls

  • Dedicated encryption key

    Every client receives an AES-256 key provisioned at onboarding.

  • TLS 1.2+ in transit

    Every request encrypted from browser to backend.

  • Tenant isolation

    Scope enforced at identity, storage, search, encryption, and AI run layers.

  • AI guardrails

    Read-only, no raw access, no secrets, no cross-tenant data, no training on your data, PII redacted.

  • Audit trail

    Every run is recorded with tools used and safety checks applied.

  • Client-controlled exit

    Export, then key destruction — ciphertext becomes unreadable, including in backups.

Ready for a walkthrough?

See Becide against your data landscape — from the pack to the assistant to the audit trail.