Skip to content

AI colleague

Samira, an AI colleague for the management team

Ask a question in plain language and get the figure, a chart and where it came from. Samira works on the same certified metrics as your dashboards, with your access rights.

Today Samira runs on demonstration data. Work on a company’s own data begins with a pilot.

What changed in sales this week? Group sales grew by 4.2%. One event: returns at Company B are three times the usual level.

Interface example. Demo data.

What Samira does

  • Answers questions

    How did revenue move this week, and why? Samira returns the number, a chart and the source metric.

  • Morning brief

    What changed overnight across the group’s companies, in a few lines.

  • Watches for events

    When a metric leaves its expected range, Samira flags it and explains what drove it.

  • Drafts for approval

    The weekly review, a summary for the board or a task for a team: Samira prepares, a person approves.

  • Respects roles

    A director of one company sees only that company. The group CEO sees the whole picture.

  • Voice

    A question asked aloud gets an answer aloud, in English and Russian today.

How a question is answered

  1. A manager asks a question.

  2. A router weighs how complex and how sensitive the question is, and chooses the model.

  3. The model receives the question and the descriptions of metrics, not the data itself, and decides what to calculate.

  4. The calculation runs in the warehouse inside the country, with that manager’s access rights.

  5. The model receives aggregates only. Names and contacts are replaced with codes and restored inside the country.

  6. Before the answer is shown, every number in it is checked against the calculation.

  7. Any outside action, such as sending an email or creating a task, runs only after a person confirms it.

Rules Samira follows

Languages

Text questions work in English, Russian and Azerbaijani. Voice works in English and Russian; Azerbaijani voice is added once its quality passes tests with real calls.

  • Samira always says that she is an AI.
  • She is allowed to say “I don’t know”.
  • Text from emails, documents and websites is treated as data, never as instructions.
  • Data sent to models through their API is not used to train them.
  • Questions that need rows with personal data are handled by an open model inside the country.
  • Before every release, Samira is tested on a fixed set of questions with known answers.

Questions

Can Samira make up a number?

Samira does not read raw tables. She calculates through the certified metric layer, every number is checked against the calculation before it is shown, and she may answer “I don’t know”.

Which language model does Samira use?

The one that fits the task and the data rules: strong commercial models under enterprise terms for analysis, and open models hosted in the country when a question touches personal data.

Can Samira send emails or change data?

Only after a person confirms the action. Samira prepares; people approve.

Start with one question

A pilot takes four weeks, covers one question your leader cares about and has a written measure of success.