In almost every company the data exists and the decision waits. It waits for a ticket, a BI model, someone with SQL who has a slot. Ask Intelligence does not promise magic: it promises that a CFO or ops lead’s question reaches the right store, runs with the role’s permission and comes back as a traceable answer. The engine is Corebrain. The data is not copied into a lake “just in case”. It is queried at the source.
Why SQL became an internal monopoly
Every new report is an engineering bottleneck. The data team becomes a helpdesk. The business learns not to ask, or to export a CSV and “fix it” in sheets. The result is double: decisions on dead data and a technical team that does not build product because it fills ad hoc requests.
Classic BI solves the known dashboard. It does not solve the question that was not in the backlog. That is where a language model, badly placed, becomes a risk: it hallucinates, mixes schemas, takes data out of perimeter. Ask Intelligence exists for that question — with route, permission and a log.
Three parts, one flow
The user asks in the workspace (“margin by product this quarter”). Corebrain parses intent, picks the source (say sales Postgres) and executes. Ask Intelligence delivers the result where the team works, not in a lab. If the source is not connected or the role cannot reach it, there is no “invent the number”: there is a no and a reason.
Etedata, when the case needs it, adds the analytics layer and chat over several sources. It is not mandatory. Some clients only switch on Ask Intelligence over two stores. Some orchestrate the ecosystem with Corebrain via SDK inside their own product. The pattern is the same: NL → route → execution at source.
Governance, or you do not ship it in a regulated firm
A Fintech cannot afford an assistant that sees every table. Permissions are the tenant’s and the role’s. The query is logged. Data does not leave for an opaque fine-tune. If your framework wants an audit trail, this is not an extra: it is the design.
That is why we do not sell “ChatGPT on the ERP” as a slogan. We sell a pipeline a CISO can explain. The diagnostic maps sources, use cases and security requirements before anything is connected.
Where value shows in the first month
Finance stops asking for “the variance spreadsheet”. Sales stops waiting for Monday’s funnel. Leadership stops briefing three people for one number in a meeting. The composite cases — finance, sales, e-commerce, ops, HR, executive briefing — live on the use-case page. None carry a client name. All describe a question that already exists in your Slack.
What you will not see in month one: a new warehouse or a team of six analysts. If that is what you want, there are integrators for it. If you want the question to stop waiting, the path is shorter.
When it is not enough
If there is no source of truth, natural language amplifies the mess. If the schema is a junk drawer with no business names, Corebrain will need context — and time. If the question is a twenty-page statistical model, it is not an Ask Intelligence case: it is data science.
The test is simple: does the answer live in one or more stores you already run? Would a human with SQL get it in minutes or hours, not a project? Then the SQL toll is the problem, not the absence of data.