Revan Analytics by DataMaaS

How your data is read

Every question runs through one secure pipeline — and you only ever talk to Revan.

The pipeline

Not guessing — reading your real, governed data.

Every question runs through one secure pipeline. You ask at stage 02 and read the answer at stage 05; the three that carry the security sit in between.

01

Your Data FoundationSales, financials, supply chain, operations and HR — consolidated by DataMaaS into one governed source rather than five you’d otherwise stitch together by hand.
02

Ask RevanYou pose the question in plain English, inside the DataMaaS Platform. This is the only stage you interact with.
03

Your governed data layerThe DataMaaS layer that holds your definitions. It holds the tools Revan is allowed to call, and it knows your schema, your KPI definitions and your business rules.
04

Secure QueryYour governed data layer generates the read-only SQL and enforces what your role may see — before a single row returns.
05

The analysisRevan reads every relevant source, establishes which basis the question needs, and returns the call. Not a raw table.

Revan itself runs on a frontier model reached through the Anthropic API, inside the DataMaaS Platform. Your question and the figures that come back pass through it. Your credentials never do — and where a result set is large, the underlying rows need not either.

The boundary

What Revan never does.

The limits matter as much as the capability — and they are enforced by architecture, not by asking the model to behave.

Never writes SQL Revan cannot emit SQL. Queries are submitted as a validated structure, never as SQL text, and your governed data layer compiles them into read-only, parameterized SQL on its own side. Injection through the model is not a policy we enforce — it is not expressible.

Never touches your database The model holds no credential and has no connection to your data. It cannot issue a query — only request that one be run.

Never sees what you cannot Excluded data does not exist for you: never returned at discovery, and blocked at execution if it is queried anyway. Revan cannot detect what was filtered out — it does not know there was anything to ask for.

Reads public sources, never feeds them Revan can look up current external facts when a question needs them. Your governed data is not part of that lookup.

Can keep large results out of the model entirely Where a result set is large, your governed data layer can hand it to a sandboxed code environment as a file rather than streaming it into the model’s context — so the underlying rows need never pass through it at all.

Access is enforced at the database — not by trusting the AI.

The governed data layer

It knows your business, because your definitions are written into it.

Generic AI tools know nothing about your business. Your governed data layer does. DataMaaS encodes your environment into it, so a question is interpreted the way your business would interpret it rather than the way a model guesses.

Your schemaYour tables, your joins, and what each field actually means — rather than a model’s guess from a column name.
Your KPIsMargin, churn, on-time delivery — defined the way your business defines them, not the way a model assumes.
Your business rulesWhat counts, what is excluded, and which basis a figure is reported on when more than one exists.

It does not matter what you run. One ERP or four in four countries and four currencies; a data warehouse, a lake, or a reporting database. If DataMaaS can reach it, it can sit behind a governed data layer — and figures can be aligned to a single consolidated currency before Revan reads them.

Two ways in

You do not need a governed data layer to start.

Point Revan at governed data, or at a file of your own. Both work — and the difference in how each is governed is worth stating plainly.

Route one Your governed data layer

Governed by your role. Table, row and column entitlements are applied before any result returns, on every query, without exception.

Route two A spreadsheet you upload

Upload a file to Revan — a large one is fine. It is governed by what you chose to put in it rather than by your role, and it stays available until you delete it.

Access control

Each user only ever sees what their role permits.

Three layers of entitlement govern every query, applied before results are returned.

Tables — what exists for youYour role sets which tables exist at all. A finance leader sees the cost and margin tables; a rep sees sales only.
Rows — security filtersDefine a filter such as “Site” once. A user scoped to North America never sees the other sites.
Columns — sensitive fieldsCost and margin are gated by role — a rep sees top-line only.

Assurance

Read-only, under your control, and never used to train AI.

Read-only, two layers Read operations only, over a read-only connection. Writes are impossible at both layers.

Never used to train AI Revan reaches Claude via the Anthropic API under Anthropic’s Commercial Terms. Your prompts and outputs are contractually excluded from model training — not a setting you switch on, and not something that can be changed for you.

You control what is kept Your questions, chats and uploaded files live in the DataMaaS Platform for as long as you keep them, and are deleted when you delete them. Separately, Anthropic deletes API inputs and outputs from its own systems within 30 days — automatic, and independent of us.

SOC 2 Type II The DataMaaS Platform is SOC 2 Type II attested, and Revan runs inside that scope.

Every query runs under your identity and your role’s entitlements.

Need more than a web page? Your DataMaaS contact can provide the SOC 2 Type II report and a security overview under NDA.

That’s what Revan can read. Next: what it does with it →

You only ever talk to Revan.

Everything above happens behind one question in plain English. The pipeline is the reason the answer can be trusted — but you never have to operate it, and there is nothing to configure before you ask.

Where next

Keep going.

Judgment How Revan thinks → The ten reflex lenses — the angles every answer is stress-tested from before it reaches you, and how ten views resolve into one recommendation. Overview Back to the overview → What Revan is, the five standing commitments behind every answer, and the shape a Revan answer always takes.

Ready to see it on your own numbers?

Bring a live question, a demo layer, or a spreadsheet of your own.

Ask Revan →