We build intelligence that works in Persian.

Bespar Fanavaran Hooshmand makes software that understands how an organisation works — its meetings, its decisions, its language — and does the work that follows. Persian comes first: right to left, Jalali dates, and speech the way people actually speak it.

Persian from the first line

Not a translation of an English product. Text, speech and the calendar start in Persian; English follows.

Our own models

We train our own speech recognition on Persian as it is spoken, from data our own team labels.

People keep the decision

Our agents act only with the asker's permission, and ask before anything that cannot be undone.

Bespar Meet

Meetings that record themselves.

Record at the table, on your phone, or upload the file afterwards. Bespar Meet writes it down as it is said, knows who is speaking, and turns the meeting into work.

1. Before

   Echo prepares each host: the last minutes, the open tasks, and what was already decided.

2. During

   A live transcript in Persian, English or both; every speaker named by the voice they recorded once; a card the moment something was decided before.

3. After

   A summary on your own letterhead, minutes the people in the room sign, and the decisions, tasks and risks on the board.

Bespar Meet

One platform. Many minds.

One sign-in and one security wall, with the working day inside it: rooms with the agents in them, a board where a project is an order, meetings that end in signed minutes, mail that arrives already drafted. Each organisation decides who may do what, and which language model thinks for it.

Bespar Label

Our own ears.

Speech recognition that understands Persian as it is spoken needs Persian as it is spoken. Bespar Label is where our team hears it, writes it down and checks it — the data our own models learn from. It is an internal tool, and we keep its details to ourselves.

Persian, as people speak it

Real speech, not read-aloud text.

Written and checked by people

Every line heard, written and reviewed by our team.

Our models learn from it

The speech recognition inside our apps is trained on it.

How we build.

Every Bespar app stands on the same few rules — the ones a team would want from a colleague it trusts.

• Security in the structure

  Access is enforced in the database, not in a prompt. An agent borrows its user's authority, never more, and cannot delete.

• Persian first

  Right to left, Persian digits and Jalali dates — built in, not translated.

• Our own models

  Speech recognition trained on our own Persian data, and the language model the organisation chooses.

• A person's yes

  Every write asks first and names what it will touch. Nothing irreversible happens on its own.

Work does itself.

You keep the last word.

© 2026 Bespar Fanavaran Hooshmand