/ 01 — BI, redrawn
Ask Azul.
Like a coworker.
Azul is an agentic analyst that queries your databases, your business tools, your documents and your files — then cross-references, synthesizes, and reads the answer in light of your KPIs.
v0.1 · Private beta askazul.fr Available Q2 2026
/ Your entire data estate
Your databases / 9
Postgres ◆ MySQL ◆ Oracle ◆ MS SQL Server ◆ Snowflake ◆ BigQuery ◆ Redshift ◆ Databricks ◆ ClickHouse ◆ Postgres ◆ MySQL ◆ Oracle ◆ MS SQL Server ◆ Snowflake ◆ BigQuery ◆ Redshift ◆ Databricks ◆ ClickHouse ◆ Postgres ◆ MySQL ◆ Oracle ◆ MS SQL Server ◆ Snowflake ◆ BigQuery ◆ Redshift ◆ Databricks ◆ ClickHouse ◆ Postgres ◆ MySQL ◆ Oracle ◆ MS SQL Server ◆ Snowflake ◆ BigQuery ◆ Redshift ◆ Databricks ◆ ClickHouse ◆
Your tools / 10
Salesforce ◆ HubSpot ◆ Zendesk ◆ Intercom ◆ Stripe ◆ Notion ◆ Linear ◆ Jira ◆ Slack ◆ Google Workspace ◆ Salesforce ◆ HubSpot ◆ Zendesk ◆ Intercom ◆ Stripe ◆ Notion ◆ Linear ◆ Jira ◆ Slack ◆ Google Workspace ◆ Salesforce ◆ HubSpot ◆ Zendesk ◆ Intercom ◆ Stripe ◆ Notion ◆ Linear ◆ Jira ◆ Slack ◆ Google Workspace ◆ Salesforce ◆ HubSpot ◆ Zendesk ◆ Intercom ◆ Stripe ◆ Notion ◆ Linear ◆ Jira ◆ Slack ◆ Google Workspace ◆
Your files / 6
Excel ◆ Google Sheets ◆ CSV ◆ PDF ◆ Word ◆ Markdown ◆ Excel ◆ Google Sheets ◆ CSV ◆ PDF ◆ Word ◆ Markdown ◆ Excel ◆ Google Sheets ◆ CSV ◆ PDF ◆ Word ◆ Markdown ◆ Excel ◆ Google Sheets ◆ CSV ◆ PDF ◆ Word ◆ Markdown ◆
Plug in your sources.
Databases, SaaS tools, business documents, spreadsheets. Azul reads your schema, your glossary, your vocabulary — and learns what matters to you.
Ask the question.
The agent breaks it down, queries multiple sources, cross-references, loops if needed. You follow its reasoning at every step.
Read it against your KPIs.
Azul doesn't just return a number. It comments, contextualizes, compares to your targets. Pin, share, replay.
One question. One answer. Not three days of waiting.
Ask a question. Azul queries your databases, cross-references your documents, calls your business tools — loops if needed, draws the charts that answer, and reads them in light of your KPIs.
azul · agentic loop 3.4s · 5 tools · 6 rows · 1 KPI
$ MRR by acquisition cohort vs Q1 OKR
agent trace 5 étapes · multi-source - ▶ azul.plan ─ break down into 3 sub-questions
- ▶ azul.query ─ postgres → public.subscriptions
- ▶ azul.fetch ─ salesforce → opportunities (closed_won, 6mo)
- ▶ azul.read ─ docs/kpi-glossary.md § "MRR by cohort"
- ▶ azul.synthesize ─ cross-reference + read vs Q1 OKR
WITH cohort AS (
SELECT date_trunc('month', s.start_at) AS m,
SUM(s.mrr_cents) / 100 AS mrr
FROM subscriptions s
WHERE s.start_at >= now() - interval '6 months'
GROUP BY 1
)
SELECT m, mrr, mrr / nullif(target, 0) AS vs_okr
FROM cohort JOIN okr_q1 USING (m);
RENDERED · BARS MRR / cohort · €
agent trace 5 étapes · multi-source - ▶ azul.plan ─ 3 signals: usage, support, ARR
- ▶ azul.query ─ postgres → usage_events (90d)
- ▶ azul.fetch ─ zendesk → open tickets per account
- ▶ azul.fetch ─ hubspot → ARR + segment per account
- ▶ azul.synthesize ─ weighted risk score
WITH drop AS (
SELECT customer_id,
AVG(events) FILTER (WHERE d > now() - '30d')
- AVG(events) FILTER (WHERE d > now() - '90d') AS delta
FROM usage_events GROUP BY 1
)
SELECT c.name, d.delta, t.open_tickets, a.arr
FROM drop d JOIN tickets t USING (customer_id)
JOIN accounts a USING (customer_id)
ORDER BY d.delta ASC LIMIT 10;
RENDERED · BARS risk score · 0–100
agent trace 5 étapes · multi-source - ▶ azul.plan ─ glossary definition + join
- ▶ azul.read ─ docs/finance/gross-margin.md
- ▶ azul.query ─ postgres → invoices + cogs
- ▶ azul.fetch ─ stripe → payouts (network fees)
- ▶ azul.synthesize ─ gross margin per segment
SELECT a.segment,
SUM(i.amount - c.cogs - s.fees) /
NULLIF(SUM(i.amount), 0) AS gross_margin
FROM invoices i
JOIN cogs c USING (invoice_id)
JOIN stripe.fees s USING (charge_id)
JOIN accounts a USING (customer_id)
GROUP BY 1
ORDER BY 1;
RENDERED · BARS gross margin · %
agent trace 5 étapes · multi-source - ▶ azul.plan ─ probability × amount × stage
- ▶ azul.fetch ─ salesforce → opportunities (open)
- ▶ azul.read ─ docs/sales/stage-probability.md
- ▶ azul.query ─ postgres → owners + quotas
- ▶ azul.synthesize ─ weighting + read vs quota
SELECT o.stage,
SUM(o.amount * p.probability) AS weighted
FROM sf.opportunities o
JOIN stage_probabilities p USING (stage)
WHERE o.is_closed = false
GROUP BY 1
ORDER BY p.probability;
RENDERED · BARS weighted pipeline · €
agent trace 5 étapes · multi-source - ▶ azul.plan ─ SLA + support channels
- ▶ azul.fetch ─ zendesk → resolved tickets (90d)
- ▶ azul.fetch ─ intercom → closed conversations
- ▶ azul.read ─ docs/sla-targets.md
- ▶ azul.synthesize ─ median time vs SLA, per month
SELECT date_trunc('month', closed_at) AS m,
AVG(EXTRACT(epoch FROM closed_at - opened_at)) / 3600
AS avg_resolution_h
FROM support.tickets
WHERE status = 'resolved'
AND closed_at >= '2026-01-01'
GROUP BY 1 ORDER BY 1;
RENDERED · BARS resolution · hours
For teams traditional BI has let down.
If you have ever waited three days for a chart, asked the analyst a favor for a small question, or dropped a question because the answer crossed too many tools — you are in the right place.
No proprietary DSL. No rewriting the warehouse. Just an analyst that speaks to all your data — the kind that sleeps in your databases, your tools, your documents.
- /01 Founders who want to dig into their own numbers without interrupting the data team.
- /02 Product teams who need answers now, not a ticket in the BI backlog.
- /03 Analysts who want to automate the 80% of repeat questions and free up time for the real work.
- /04 Finance leads who want a clear audit trail on every number — Azul always shows its sources.
/ 05 — Guarantees
Four commitments, not promises.
Before plugging AI into your data, here's what you can count on.
Read-only, at the source
Azul connects to your systems with read-only credentials. The ability to modify isn't disabled — it doesn't exist.
Your data never trains a model
No customer data — schemas, queries, answers — is ever used to train or improve any model, ours or a third party's.
Hosted in Europe, or on your servers
EU infrastructure by default. On-premise deployment available — Azul runs in your VPC, your data never leaves it.
Every answer is traceable
Every query, every API call, every document consulted is logged alongside the agent's reasoning. Any past answer can be reopened and its sources verified.
/ 06
See Azul on your own data.
Thirty minutes on a call. Bring a schema, an API, a document — we plug in, we ask questions together.