This is the end-to-end roadmap: from creating the account to a shared dashboard and a question answered by AI. Each stage has a dedicated page — here is the map and how they connect.
Why we don't promise a time
The effort of each stage varies with the volume of your data, the source, the provisioning, and external authorizations (database access, the ERP's API key). We don't measure an end-to-end time, so we don't publish one. This page's address keeps the old name so saved links don't break.
All examples use the fictional workspace Comércio Aurora Ltda (workspace
Aurora Varejo, dataset prefix aurora).
Prerequisites#
- An email for the account and an authenticator app on your phone (for the required MFA).
- A sample file (CSV/Excel) or a source you want to connect.
- Nothing else: you don't need to set up any cloud infrastructure.
The timed roadmap#
Stage 1 — Account, MFA, and datalake
Create the account and go through the onboarding wizard (5 steps: Company → Plan →
Security/MFA → Datalake → Finish). At the end, the three medallion datasets
(prd_aurora_bronze, prd_aurora_silver, prd_aurora_gold) are created and you land on the
overview dashboard.
→ Onboarding: the wizard step by step · Account, organization, and workspace
Stage 2 — Connect your first source
Register the connection in ETL/ELT → Connections with New connection and use Test connection: the result is split into Network, Authentication, and Read. Saving a connection doesn't start any ingestion.
Stage 3 — First ingestion (publish and run)
Build the pipeline in ETL/ELT → Pipelines and use Publish (schedule off): this creates version 1 and runs nothing. Then Run now runs the first load and Turn on schedule turns on repetition — they are three distinct actions.
→ First ingestion · Publish and version
Stage 4 — View and build dashboards
Opening a dashboard puts you in Viewing; Edit is an explicit action. To create a new dashboard you must enter a Dashboard name and choose under How do you want to start? between From a template, Use my data, and Blank.
→ View a dashboard · Build your first dashboard
Stage 5 — Ask your data (AI)
Open the question wizard, choose the scope (a Gold table or the semantic model), and ask in plain language. The AI generates the SQL with a cost guard and answers with provenance — or declines if it has no basis.
Stage 6 — Share
Publish a secret link for the leadership team and schedule an email subscription for the managers — each one receives only the rows they're allowed to see (RLS on delivery).
→ Share the dashboard · Public links and snapshots
After the first journey#
With the basics running, these screens become part of your day to day:
- Overview dashboard (
/overview) — the workspace summary (sources, dashboards, forecast cost). - Lineage (
/linhagem) — where each number came from and what breaks if you change a table. - Costs and forecast (
/costs) — the monthly estimate of subscription + usage. - Audit log (
/auditoria) — who did what, when, and how much it cost.
Expected result#
At the end you'll have: a provisioned datalake, a source ingested into three layers, a dashboard published through a secure link, an email subscription scheduled with RLS, and at least one question answered by AI with provenance.
Limits and costs#
- No time is promised: actual ingestion depends on the volume of your data, the source, and the run queue.
- Cloud usage (processing + AI) is paid in credits, with a cost guard. Registering an account/source/dashboard costs R$ 0.
- Availability: 99.5% monthly SLA from the Growth plan (see /legal/sla); follow it at /status.
Common errors#
- Empty dashboard in Stage 4: Gold hasn't been materialized yet — check the Data: label on the dashboard; if it says no data yet, run the load first.
- Action blocked (402): subscription past due; settle it in Plans (expensive actions are blocked).
- AI declines the question: it has no basis in the model — rephrase or ground it in the semantic model.
Related#
Last reviewed: 2026-09-16.