AI assistant for your company's data
A brain
with hands.
Asking your data is something half the market does now. BRAIN answers, leaves the dashboard built and then acts: it hands out the task, fires the process your company already has, and reports back on how it's going. It lives inside your database, read-only.
-- written just now; it wasn't saved anywhere SELECT d.name, SUM(v.amount) AS sales FROM sales v JOIN distributors d ON d.id = v.distributor_id WHERE d.region = 'north' AND v.date BETWEEN '2026-08-01' AND '2026-08-31' GROUP BY 1 ORDER BY sales DESC;
| Distributor | Sales | MoM chg. |
|---|---|---|
| Pacífico | S/ 501 400 | −3 % |
| Norte SAC | S/ 412 800 | −31 % |
| Andina EIRL | S/ 298 100 | −24 % |
| Selva | S/ 188 900 | +2 % |
The problem
The number arrived. The meeting happened. And nothing changed.
Almost every company has enough data to decide well. Getting the number isn't the problem any more — half the market solves that. The problem is the distance between the number and somebody actually doing something about it.
I ask Juan for the report and it reaches me the next dayThe data arrives after the decision.
We have Power BI, but only three people use itThe tool exists. The access doesn't.
We presented it to the committee and that was thatNobody left the room with a task.
By the time I find out, the month has closedFinding out late costs the same as not finding out.
What it does
It answers. It leaves it ready. And it moves it.
Almost every tool stops at the first one.
Answers
You ask in plain language and it answers: the number, why it happened, and the exact query it ran to get there. Without asking anyone for a report or waiting until tomorrow.
It leaves it ready
What came out of the chat ends up as a dashboard that refreshes itself, a deck or a spreadsheet. With the technical appendix, for whoever shows up to argue about it.
And it moves it HANDS
Sends the report every Monday at seven. Tells you when the indicator crosses. And fires the process your company already has, without anyone launching it by hand.
Answers
From «how much did we sell yesterday?» to «why is this happening to us?»
The number, right now
Writes the query, runs it and answers. No prebuilt reports.
- «How much did we sell yesterday by region?»
- «How's coverage doing?»
Why it happened
Raises hypotheses, fires several queries and rules out explanations before answering.
- «Why did the north region fall?»
- «Is it price or volume?»
What's going to happen
Projects over the history and states how confident it is.
- «How are we closing the month?»
- «Do we hit quota?»
What explains it
Ranks the factors by explanatory power, and says which ones explain nothing.
- «What sets the growing ones apart?»
What you didn't ask
Reviews the period and brings back what stands out, with how serious it is.
- «What looks odd this quarter?»
And what isn't in the database
Policies, manuals and contracts you upload, citing the source.
- «What does the discount policy say?»
Why you can trust the number
Anyone can hand you a number. Here you see where it came from.
A language model will write any conclusion with total poise, and write it well. That's why BRAIN's analyses pass checks before they reach you — and it shows you the checks, so you don't have to take its word for it.
The effects have to add up
When it decomposes a variance, it checks that the parts add up to the total and reports the residual. If it doesn't reconcile, it says so — instead of spreading the difference across the factors to make it fit.
residual S/ 0 · 0.0 %The forecast has to earn it
Several methods compete under blocked cross-validation, and the one with the lowest error wins. It's always compared against a naive benchmark: if it can't beat that, the forecast isn't worth having, and it says so.
Holt-Winters · MAPE 6.2 % vs 14.8 %A weak analysis doesn't ship
Every pipeline demands a minimum number of findings, a minimum confidence and certain kinds of conclusion. If it falls short, it reframes the analysis and tries again before answering you.
retries, it doesn't lower the barEvery claim carries its source
Conclusions carry citations, and each citation opens the query or the document behind it. Your analyst doesn't have to take anything on faith: they can redo it.
[1] … [6] → sql, kpi, document, webNot everything deserves interrupting you
Separates what it observes from what it diagnoses from what it considers an alert. A curious finding and an urgent problem don't arrive carrying the same weight.
observation · diagnosis · alertAnd what isn't in your database
A diagnosis can cross your data with your documents, your KPIs and outside context — exchange rates, weather, competitors — when the question calls for it.
sql · kpi · documents · web · mcpDelivers
The chat closes. The dashboard is still there.
Three ways to keep it. Which one you want depends on who's going to open it.
The same question, next month
You save the query with its chart, its owner and its version. Next month you open it and it runs again against today's data, without anyone having to remember how it was calculated. In the product it's called a Dive.
It re-queries every time someone opens it
It doesn't store a snapshot of the numbers: it asks for them again, and applies the permissions of whoever is looking. The manager and their supervisor open the same board and see different figures, each their own.
With the SQL in the appendix
Cover, the conclusion with its caveats beside it, one slide per finding and the recommendations ordered by priority. The queries go at the back, for whoever would rather rebuild the number than take it on faith.
HANDS
Finding out in time is worth nothing if nobody then does anything.
HANDS are BRAIN's hands: what it can do with a finding, not just report it. Schedule the report, watch the indicator, hand out the task and fire the process your company already has.
Actions
BRAIN doesn't need a connector for every system. Your company already has them.
BRAIN doesn't try to rebuild the processes you already set up. It fires. If your flow asks for an approval in Teams and waits for the manager to decide, creates the task in Planner, writes to the ERP and pings whoever is out in the field — BRAIN only has to tell it to start, and with what data.
From the conversation to the process
What this looks like in practice
BRAIN fires the process
An administrator registers the flow once: what it's called, what data it takes and where it goes. From then on BRAIN knows when to use it, and you ask for it in plain language.
The process asks BRAIN
It works the other way too. A flow already running stops, asks, and carries on based on the answer — with the permissions of the person it runs as, and nothing more.
And the brakes, because this reaches outside
How it works
Five steps between the question and the answer.
None of them is a model guessing. Each one is an agent with one concrete job, and the whole result can be audited.
read-only, with the credentials you provided. Your data is never copied anywhere.Where it lives
Control
The four questions IT always asks.
The model sees what your tables and columns are called, so it can write the query. The query runs on your database and what comes back is the aggregated result. Nothing is copied.
Three layers stop it: a validator that rejects every write, the session opened read-only, and our recommendation to connect with credentials that have no write permission.
Data filters are set per person and they intersect: they never widen, only narrow. The same bot shows a supervisor their region and the manager all of them.
Who asked what, which query ran and what was delivered. A complete, searchable audit trail.
What it still doesn't do
We'd rather tell you now than in month three.
Almost no product page has this section. We put it here because the alternative is you finding out after signing.
Frequently asked questions
What people ask us before the demo.
What is BRAIN?
BRAIN is an artificial intelligence assistant that answers questions about your company's database in plain language. It writes the SQL, runs it read-only, shows you where every number comes from and leaves the dashboard built. And if you ask it to, it acts: it creates tasks, fires your company's Power Automate flows and posts to Microsoft Teams. It is built by Vida Software.
What is a data query bot?
It's an assistant you ask about your data the way you'd ask an analyst —“how much did we sell in the north region in August?”— that turns the question into a query on your database, runs it and answers with the number. What sets one apart from another is what happens next: whether it lets you see the query, whether it checks the result against the real data, and whether it can do something with the answer. BRAIN does all three.
Can I ask my database questions in plain language, without knowing SQL?
Yes. You type the question the way you'd say it in a meeting and BRAIN writes the SQL query for you. The query is always visible: whoever knows SQL can review it, and whoever doesn't can pass it to their data team to validate.
Which databases does it work with?
PostgreSQL, MySQL, Microsoft SQL Server, DuckDB and MotherDuck. It connects with a read-only user you provide and queries your database live, instead of copying it somewhere else.
Is it safe? Can it modify or delete my data?
It can't modify them. It connects read-only and, on top of that, every query goes through a validator that rejects any statement that would write, delete or change anything before it runs. Credentials are stored encrypted, and each person sees only what belongs to them: if someone from the north region asks about sales, the answer comes back filtered to the north region.
How is it different from ChatGPT?
ChatGPT, on its own, isn't connected to your company's database: it answers with what it knows or with whatever you paste in. BRAIN queries your data live, every number comes from a query you can see and rerun, and it knows your business —your KPIs, your regions, your brands— because you set it up with them.
Does it replace Power BI or another BI tool?
No, it complements it. A BI tool shows the reports someone designed in advance; BRAIN answers the question that comes up today and that no report has, and turns it into a dashboard if you're going to ask it again. It also does what a report can't: hand out the task and fire the process.
Does it work inside Microsoft Teams?
Yes. BRAIN lives as a bot in Teams, where your team already works: they message it and it answers right there, with the same access rules as on the web. It also replies by email to approved senders, and has channels for Slack, WhatsApp and Telegram.
Can it create tasks or fire processes?
Yes. It creates tasks in Microsoft To Do, fires the Power Automate flows your company already has —approvals, WhatsApp alerts, emails— and posts to Teams. Before firing a process it shows you exactly what it's going to send and waits for your confirmation. It doesn't write directly to your ERP or CRM: if one of your flows does, the flow writes it, with its own permissions.
What do I need to get started? Do I have to install anything?
There's nothing to install on your machines. You need read-only access to your database and, to get value from day one, two or three KPIs you already track. The demo runs on your own data: you connect us read-only and you ask the questions, in front of your team.
The demo
We bring your data, not ours.
A demo on sample data proves nothing. Connect us read-only to a representative database and you ask the questions yourself in the meeting, in front of your team.