# İDRAK, the layer that knows what your data means

> İDRAK reads your tables, learns how your company speaks, refuses to guess when it is unsure and shows the source behind every answer. AI works from that map.

- Page: https://dehalab.com/en/idrak
- Language: English (en)
- Same page in Turkish: https://dehalab.com/idrak
- Markdown: https://dehalab.com/en/idrak.md
- Site index for agents: https://dehalab.com/llms.txt

## It does not know everything. It knows your business.

İDRAK holds one map of what your data means and how the pieces connect. AI works from that map rather than from the whole world. Narrowing the ground is not a compromise, it is the precondition of an AI that works, because a bounded domain is one whose behaviour you can predict.

### Map: It works out what your data means

It reads your tables, works out what each one describes and which records belong together. You never draw a schema by hand.

### Vocabulary: It learns how your company speaks

Different names for the same thing, what a value actually means, which dates last quarter covers in your own calendar. All of it is written into the map, so every question is read the same way.

### Bounds: It does not guess when it is unsure

A question that falls outside the map, or that could mean more than one thing, does not get an invented answer. The system asks you what you meant, or says it cannot answer this one.

### Evidence: Every answer carries its sources

An answer arrives with the tables and records it was built from. You do not have to trust the system to check it, you open the source.

### The machine suggests, you confirm

Every relationship the system finds on its own comes to you first. What you confirm can never be overwritten by the machine later.

### Four kinds of reasoning, one platform

Rule checks, relationships, forecasting and optimization run one after another. What the forecast produces is what the optimization consumes, without a hand-off in between.

### It sharpens as you use it

Questions that come up again turn into fixed, exact answers. Over time the system leans less on AI and more on exact computation.

### It labels, groups and flags

The system lays labels over your data, groups look-alikes and flags outliers, even from a product photo. Each signal becomes something you can query: ask for the revenue of customers labelled at risk.

### Documents join the map too

The customer in a contract and the customer record in your database are linked with a page and paragraph reference. Who signed this clause becomes a question with an answer.

### It holds meaning, not your data

İDRAK knows what your data means without keeping the data itself. Your business data stays on your own infrastructure.

**Contact:** ensar@dehalab.com · https://dehalab.com/en/contact