Skip to main content
About

About Shujing Intelligence

An AI-native enterprise software company, starting with the global-trade workspace.

Who we are

Rebuild the way work happens, then talk about AI

Putting AI into an enterprise is not the same as handing the enterprise to AI.

A global-trade team’s work is split across mailboxes, drives, chat, spreadsheets and translation tools. Material cannot be found, versions do not match, and terminology is re-confirmed every time.

We did not make AI a new entrance. We rebuilt permissions, material and actions first: AI reads as the person who asked, and a human confirms before any write.

The Dataflare AI Global Trade Workspace is the first product built this way. It solves documents, translation, mail and knowledge first, then puts those capabilities under one organization and permission model.

Why global trade

Three concrete problems

No grand narrative — start with what happens every day.

Material lives with individuals

Quotations, document templates and supplier data sit in mailboxes and personal drives; when a person leaves, the material leaves too.

Translated files are not ready to send

Generic translation scrambles tables, headers, footers and numbering, so someone has to rebuild the layout.

The mailbox is not under the company’s control

Binding a third-party mailbox gives visibility, but not ownership of the domain, delivery or data.

Our principles

Four things we do not trade away

These principles live in the product and in the boundaries of the code.

Facts before adjectives

Every number, format and capability on this site must have a counterpart in the code.

Private by default

Private deployment is supported; mail is visible only to its owner, not even to an admin.

People confirm consequential actions

AI may read and draft, but writes and outbound sends show the full content and wait for a human.

One permission model, no side doors

Roles gate feature entry; content access only follows node ACLs. Search and AI use the same rules.

How we build

Product and AI in one system

We did not ship a chat box and then go back to add permissions.

  1. 01

    Define the permission boundary

    Every capability first answers who may read, change and send. The model comes after that.

  2. 02

    Verify with real files

    docx, xlsx, legacy .xls and PDF are checked against their actual layout, not a mock-up.

  3. 03

    Make the limits visible

    Unsupported capabilities, lossy formats and scans that need OCR are written down in the product and docs.

Company details

Registration and contact details

These details follow the company’s confirmed registration record.

Confirmed registration details are shown above; unconfirmed fields are omitted.

Legal entity
苏州数鲸智能科技有限公司
Legal representative
Huang Songkang
City
Kunshan, China
Registered address
昆山市玉山镇祖冲之南路1699号综合北楼1304室
Registered capital
RMB 2,000,000
Established
April 2026
Status
Active
Company size
Micro enterprise

Want to understand what we are building?

Bring a real global-trade scenario and talk to us.

Contact us
Book a demo