The Azul product: from the question to the dashboard

From question to chart, every query is checked.

1
You ask the question
In your own words. No SQL, no data model to learn.
2
The agent writes the query
From your schema and your business context: your tables, your definitions, your rules.
3
Azul checks it, then runs it
Read-only, within the scope you chose, without weighing on your production. A query outside that frame is refused.
4
You get the answer, and dig deeper
Figures, chart, commentary. The next question keeps the thread of the conversation.

Query written by the agentPostgreSQL

SELECT c.segment,
       SUM(i.amount_excl_tax) AS revenue
FROM   invoices i
JOIN   customers c ON c.id = i.customer_id
WHERE  i.invoice_date >= DATE '2026-04-01'
  AND  i.invoice_date <  DATE '2026-10-01'
GROUP BY c.segment
ORDER BY revenue DESC;

Read-onlyno change possible
Allowed tablesinvoices, customers
Production preservedbounded duration
Run on your database4 rows

Before it answers, Azul learns your business.

An agent that ignores your definitions gives plausible answers, and wrong ones. Azul builds that context with you, once and for all, then uses it in every answer.

1
It reads your documents
Procedures, glossaries, business notes in PDF, scanned documents included.
2
It explores your schema
Tables, columns, relations, real values: it infers everything that can be inferred.
3
It asks you the right questions
Only what it cannot guess, as multiple choice or free text.

A few hours with your teams, rather than weeks of modelling workshops.

Try it: pick an answer, then send.

All your sources, one conversation.

Your databases
PostgreSQL, MySQL, Oracle. Several databases in one workspace.
Your files
CSV exports and Excel workbooks, imported with the help of AI and queryable like a table.
Your PDF documents
For business context, or to extract rows from them (invoices, quotes, statements). OCR for scans.
Your business tools
CRM, ERP, ticketing, in-house tools: any API, connected directly or through MCP.

The agent picks the right source for each question and brings the results together. You can also choose the database yourself.

Your business tools too: query, cross-reference, act.

The agent does not stop at your databases: it queries the tools where your business runs, cross-references their answers with your data, and acts in those tools when you ask it to.

Any API, with or without MCPCRM, ERP, ticketing, in-house tool: an API connects directly, or through its MCP server.
One more data sourceThe agent queries your tools and cross-references their answers with your databases, in the same conversation.
Actions in your toolsCreate a ticket, update a customer record, schedule a follow-up: only in the tools you connect, with only the actions you authorise.

For example

Salesforce
HubSpot
Zendesk
Your ERP
Your in-house tools

Your databases stay read-only. Every action in a tool is recorded in the audit log.

Which key accounts cut their orders this quarter and have an open support ticket?

Sales databaseorders
CRMaccounts, sales reps
Ticketingopen tickets

Three key accounts combine a drop in orders and an open ticket:

Account	Orders	Open ticket	Sales rep
Maison Castel	−18%	Late delivery	Claire Martin
Norvia	−12%	Billing error	Thomas Robert
Tessier et Fils	−9%	Portal access	Claire Martin

Create a follow-up task for each, assigned to their sales rep.

Action in the CRM3 follow-up tasks created, assigned to the sales reps concerned.Recorded in the audit log

The dashboard for ongoing monitoring

The useful charts, gathered and refreshed automatically: every hour, day, week or month.

The report for the one-off decision

An analysis written by the agent, with its charts frozen at a date: the document you send to the committee.

Sales review, third quarter 2026

30 September 2026

The figures below are frozen on the runs this report was written from. They are never refreshed.

Summary

From April to September, revenue grows 8.4% year on year. Growth comes first from key accounts, which now weigh 42% of the total.

€3.2MRevenue, 6 months

+8.4%year on year

42%Key accounts

1. Monthly trend

September is the best month of the period (€0.62M), after the usual August dip (€0.47M).

The alert for automatic watchfulness

On the metrics that matter, no more checking by hand: you set a threshold or a condition, and as soon as a scheduled refresh crosses it, Azul e-mails the people who follow the metric.

A threshold or a condition on the metric of your choice
Checked at every scheduled refresh
An e-mail to the people who follow the metric

Large databases?Azul queries them, it does not copy them.

Each question becomes a targeted query, run where your data lives.

In-memory BI

A copy to build, size and refresh.
The figures of the last load.
Analysis stops at what was loaded.

Azul, the query at the source

Nothing to preload, no copy to size.
The database aggregates: only the result comes back.
Current figures, for every question.

Every query is bounded in time. An analysis never drags on over your database: your production is preserved.

If your database itself is the limiting factor, we help you with our integration partner: read-only replica, indexes, sizing.

Built around AI, not retrofitted with it.

	Legacy BI with an AI assistant	General-purpose AI, ChatGPT-style	Azul
The question that has no dashboard yet	The assistant stays bound to the data model already built.	Possible, provided you copy your data into it.	Asked in plain language, directly on your databases.
Knowledge of your business	Carried by a semantic model, built as a project.	To be re-explained in every conversation, by every user.	Gathered once at onboarding, shared by everyone.
The frame given to the AI	Inherited from an architecture designed before AI.	No table scope, no validation, no roles.	Read-only, scope, validation, roles and audit, designed for an agent.
Freshness of the figures	Often a copy of the data to load and refresh.	That of the last exported file.	On-demand queries on your database: today’s figures.
Your business tools	Connectors that load their data into the model, as a project.	Plugged in account by account, with no company-wide frame.	Queried and cross-referenced with your databases; only the actions you authorise, logged.

Your BI tools remain excellent reporting tools. Azul covers what happens before and around them: the question that does not have its dashboard yet.

Every function finds its questions.

Finance and public bodies
Which budget lines exceed 90% of their allocation?
What is our average supplier payment time?
Budget tracking, variance to forecast, cash.
Alert as soon as a line exceeds 90% of its allocation.
Sales, SMBs and mid-sized companies
Which customers cut their orders by more than 20% in a year?
What margin does each product line make this quarter?
Revenue, margins, retention, team performance.
Alert when a segment’s weekly revenue drops below target.
Human resources
How has absenteeism evolved by department since January?
Which positions have been open for more than three months?
Headcount, absenteeism, turnover, hiring.
Alert when a department’s absenteeism exceeds its threshold.
Operations
Which items risk a stock-out at the current outflow rate?
Which carrier is most often late?
Inventory, lead times, quality, logistics.
Alert as soon as an item falls below its safety stock.

Next step: a pilot on your data.

The most telling thing is to watch Azul answer your own questions.

Try for free Book a demo

Or write to us at hello@askazul.fr

1
Choose a use case
A business question that matters, and has no quick answer today.
2
Prepare the data
Read-only access to the database concerned, or a file export.
3
Test Azul
Your teams ask their questions, on their data, supported end to end.

From question to chart, every query is checked.

You ask the question

The agent writes the query

Azul checks it, then runs it

You get the answer, and dig deeper

Before it answers, Azul learns your business.

It reads your documents

It explores your schema

It asks you the right questions

All your sources, one conversation.

Your databases

Your files

Your PDF documents

Your business tools

Your business tools too: query, cross-reference, act.

Three ways to put an answer to work.

The dashboard for ongoing monitoring

The report for the one-off decision

The alert for automatic watchfulness

Large databases?Azul queries them, it does not copy them.

In-memory BI

Azul, the query at the source

Built around AI, not retrofitted with it.

Every function finds its questions.

Finance and public bodies

Sales, SMBs and mid-sized companies

Human resources

Operations

Next step: a pilot on your data.

Choose a use case

Prepare the data

Test Azul