Voice input in 1C: tasks, documents and resolutions by voice

Tasks, assignments and resolutions are still typed by hand, and after a meeting someone sorts the notes into items manually. The assistant takes them by voice: you say “create a task for Sidorov to approve the invoice by 17:00”, and it recognizes the speech locally and prepares an entry in 1C. It turns meeting minutes into a task list by itself.

Discuss your task WhatsApp

What slows down input

Tasks and assignments are typed by hand: that is inconvenient while driving or on the go, and some simply get lost.
After a meeting someone replays the recording by hand and breaks it down into tasks.
A resolution on a document is easier to dictate than to type, but there is nowhere to dictate it.
Online dictation through the cloud cannot be allowed near amounts and IINs: the data would leave your perimeter.

What it does

Takes tasks by voice with an assignee and deadline: “Sidorov, approve the invoice by 17:00”.
Creates documents and resolutions from dictation.
Recognizes speech locally with Whisper technology, with about 95% accuracy.
Works without the cloud: audio and text do not leave your perimeter.
Transcribes meeting minutes and extracts tasks from them automatically.
Creates entries as drafts: posting and assignment happen only after confirmation.
Understands business speech in Russian, including names and amounts.

How it works4 steps

You say a command

“Create a task for Sidorov to approve the invoice by 17:00”, or you dictate a resolution or document text, from a phone or headset.
Speech is recognized locally

A model based on Whisper technology converts voice to text right in your perimeter, with about 95% accuracy: audio does not go to the cloud.
A draft is put together in 1C

The assistant understands what you meant (a task, a document or a resolution) and creates the right entry with an assignee and deadline.
You confirm

The entry arrives as a draft for confirmation. Nothing is written to the database until you say “OK”, as in all our scenarios.

Result

≈ 95% accuracy of local speech recognition

No cloud audio and text stay in your perimeter

Meeting → tasks the minutes turn into a task list by themselves

Who it is for

Managers who find it easier to dictate a task than to type it.
Companies where meeting notes are broken down into assignments by hand.
Businesses that need voice and amounts to stay out of cloud services.

Frequently asked questions3

How accurate is speech recognition?

A local model based on Whisper technology gives about 95% accuracy on business speech. Uncertain parts are highlighted in the draft and can be fixed in a couple of seconds before confirmation.

Does the voice go to the cloud, as with Alice or “Hello, 1C”?

No. Unlike cloud voice assistants, recognition runs locally in your perimeter. Audio and transcripts do not leave the company, so names, amounts and IINs stay inside.

Is the entry posted in 1C right away?

No. A voice command is turned into a draft and waits for your confirmation. As in all our other scenarios, nothing is written to the database without an “OK”.