Every IT system,
a workplace of AI employees
that think and act.
Biz Assistant adds a small "bridge" to business systems that used to just wait for create, read and update. AI employees come alive and, in each employee's own chat, notify, ask, draft and remember. No rewrite of business logic. People always decide.
LEGACY SYSTEMCRUD
| Order # | Status | Ship by |
|---|---|---|
| A-10231 | Received | 11/12 |
| A-10232 | Received | 11/12 ⚠ |
| A-10238 | Pending | 11/12 ⚠ |
| A-10240 | Shipped | 11/10 |
| A-10244 | Received | 11/11 ⚠ |
v1
3 orders are past their ship-by date.
▶ Please check the ship dates on the Orders screen.
Check
Approval has been waiting for over 2 hours. I have asked the person in charge.
What is Biz Assistant Framework?
An open-source framework that turns an existing or new business system into a workplace of AI employees. Add a small bridge (Bridge v1) and AI employees notify, ask, draft and remember in each staff member's own chat. People always decide.
| What it does | Tells the right staff about business changes, requests approvals, drafts documents, reports on schedule and remembers decisions |
| Where | In each staff member's own chat (12 platforms including Slack, LINE, Microsoft Teams, WeCom and Lark) |
| Safety principle | The AI only proposes. People confirm, approve and pay in the business system. It reads only what the user may read |
| How to connect | No rewrite of business logic. Add read and propose endpoints, a manifest and recipient rules; verify with the 52-check conformance suite |
| Licence | Apache License 2.0 (open source, free for commercial use). Source: ITC-cloud-soft/biz-ai on GitHub |
| Provided by | ITCSoft Co., Ltd. (Tokyo), which also offers adoption support and the sister product "Ops AI Bot" for AI operations and maintenance |
IT goes from "a tool that records"
to "a colleague who works with you".
The information is already in your system. What was missing was a way to bring it to people and get things moving.
The era of records
Ledgers and input. People type data in, people search for it.
The era of screens
CRUD and business logic. People open screens, check, and chase each other.
The era of intelligence
AI employees watch, notify, ask, draft and remember. People decide in chat.
BEFORE — systems that wait
- Nothing happens until someone opens a screen
- Important changes go unnoticed unless you look
- You chase people for approvals and submissions
- You collect data to write reports
AFTER — systems that think and act
- AI employees notice changes first
- Only the people concerned are told, in their own chat
- Approvals take one tap in chat
- A summary arrives every morning and evening
Think. Remember. Act on their own.
An AI employee is an AI assistant that works within each employee's own rights.
Think
Reads business data within the user's rights, spots what matters now, and drafts plans and corrections.
Remember
Remembers each person's preferences and open items, the organisation's procedures and earlier conversations. Never personal data.
Act on their own
Watches for changes, tells only those concerned, asks the right person, follows up, and sums up on schedule.
The AI reads and proposes. Confirming, approving and deciding amounts is always up to people.
A day with your AI employee.
2 items need action, 1 for reference.
The job planned for 09:00 has not started yet.
Please confirm October billing.
Approve
2 AI proposals. Please accept or decline.
All screens are fictional examples.
Your business system stays the source of truth.
AI employees work across the bridge.
The business plane and the AI plane are joined by Bridge v1, a small versioned HTTP protocol, and a machine-readable self-description, the manifest.
🏢 Business system
- Business data · staff & roles
- Rules · approvals · amounts · audit
- Who gets notified
Bridge v1
🤖 AI plane
- Conversations · long-term memory
- Automatic tasks · scheduled reports
- 9 plugins
💬 Chat
- Slack
- LINE
- Teams and more
Business changes are recorded in the same transaction
Your system decides who is told
The AI employee reads with the user's rights and drafts proposals
People decide in chat; results are recorded in your system
Safety enforced by
code and configuration, not prompts.
People decide
The AI only proposes writes. Accepting and approving is done by a person, once.
Never beyond the user
AI rights = what the AI is allowed ∩ that employee's role. Tokens live 5 minutes at most.
The system picks recipients
The AI never chooses who receives anything. Only the people concerned get it.
No personal data in transit
Events carry IDs, dates and counts only. Text is screened before sending, or fixed.
Closed by default
Group chats, unattended work and delivery targets are closed by default.
Verified by 52 checks
Promises are proven by the conformance suite. Releases need 52 / 52, nothing skipped.
For legacy systems and new ones alike.
No rewrite of business logic — but it does not connect with zero changes either. The first assessment lists what to add.
What you add: a "bridge"
- An API to read business data and to receive AI proposals
- A manifest (the system describes itself)
- Each employee links their own chat — with your system's own login
- Recipient rules decided by your system
- Events written in the same transaction as the change
- Approval rules · scheduled reports (counts only)
Born to work with AI employees
- Build on server-kit and the bridge comes with it
- You write only the business, hooks, manifest and screens
- A sample business system shows the layout and style
const bridge = await createBridgeApp({ manifest: { file: './bridge.manifest.yaml' }, store, serviceToken, bffSecret, hooks: { employees, authenticate, isSessionLive, recipientsFor, approvals, proposalKinds, reports, piiScan }, })
Keep the code, put the bridge outside
- Logins can be verified from outside (OIDC / SAML)
- APIs can be called as the user
- Changes are announced after commit (webhook / CDC)
- Approval and write APIs exist
The spec is OpenAPI. The promises are tests.
Describe the system in a manifest, wire the hooks, prove it with the 52-check conformance suite. The source is on GitHub, and an integration skill for AI coding agents is included.
bridge: 1 system: { id: order-desk, name: Order Desk } roles: manager: { delegation: propose } staff: { delegation: read } events: order.overdue: payload: [orderId, dueDate, count] # IDs, dates, counts only notify: { template: order.overdue } tools: - name: order_list_overdue http: GET /api/orders/overdue access: read roles: [manager, staff]
$ node --test packages/protocol/conformance/http.test.mjs ✔ P1–P6 protocol hygiene ✔ L1–L10 link yourself only · B1–B4 personal bots ✔ D1–D7 delegation = allow-list ∩ role ✔ S1–S5 assistant tickets bound to the session ✔ PR1–PR3 · A1–A4 AI proposes, people decide ✔ E1–E4 · N1–N3 · R1–R3 · M1–M3 no personal data 52 passed, 0 known gaps, 0 skipped
The chat your team already uses becomes the AI employee's workplace.
Blue frame: organisation bots / all: personal bots each employee connects / email is notification-only
And
the system grows itself.
Ops AI Bot watches over, fixes and grows the system itself: it takes issues, investigates, fixes and tests, opens a PR, releases after approval and remembers what it learned. Production always needs a human approval.
grows with use
Frequently asked questions
Can we keep our existing system's code unchanged?
Business logic does not need rewriting, but a bridge must be added (read and propose APIs, the manifest, chat linking, recipient rules, event recording…). The first assessment lists what to add and how much.
Could the AI approve or pay on its own?
It cannot. The AI only proposes; confirming, approving, paying and deciding amounts are done by people in the business system. The protocol and its tests prove it.
Which chats are supported?
Slack, LINE, Microsoft Teams, WeCom, WeChat, Lark and more. Personal bots that employees connect themselves work on 12 chat platforms. Email is used for notifications only.
How is personal data handled?
Events and notifications carry only IDs, dates and counts, and text is screened before sending. The AI employee's memory keeps no personal data either.
Take your IT
into the next era.
Start with one business process. We size what to add and grow it, proven by the conformance suite.