Fragmented Information:Materials, tasks, cases, and reminders were scattered across multiple tools.
← Back to All Projects02 / 04
AI Workflows / Business Systematisation
AI-Assisted Business Operations Infrastructure
From Fragmented Workflows to a Traceable System
Project Type:Work-in-Progress Operations Infrastructure
Business Context:Day-to-day collaboration within a family-law services team
Current Stage:Core design completed; end-to-end workflow closure still being developed
Submodules
01
Feishu Business Hub:Brings inputs, cases, files, tasks, outputs, and reminders into a unified view of business status.
02Case-Library Production Agent:Connects case content, templates, supporting materials, and quality checks into a reusable production workflow.
03Agent-State Governance:Uses states, stage gates, logs, and human authorisation to manage risk in AI-assisted collaboration.
System Structure
Work OS Inbox
↓Client-Case Master Table
Files and Materials
Subtasks
Output Tracking
Recurring Tasks
Project Background
The problem was not a lack of ability to do the work:workflows were fragmented, status was difficult to track, and experience was hard to reuse.
Unclear Status:It was difficult to see at a glance what needed to happen next in any given case.
Outputs Were Difficult to Retain:Case libraries, working methods, customer personas, and review notes often remained as work that had been completed but never formally captured.
AI Was Difficult to Embed in Real Workflows:AI could generate content, but without a workflow to receive and advance that output, it remained one-off text.
Build Path
The work began with identifying the underlying problems:only then were the workflows connected to Feishu and AI.
Identify the Problem:Work was scattered across folders, chat histories, personal memory, and temporary spreadsheets.
Break Down the Business Objects:The workflow was divided into inputs, cases, files, tasks, outputs, and reminders.
Build the Hub:Feishu Base was used to hold and track business status.
Introduce AI:AI was used to generate candidates, organise information, and support workflow decisions.
Execute with Controls:Dry runs, human confirmation, and stage gates were used to reduce execution risk.
Capture and Reuse:Cases, methods, personas, and review findings were brought into reusable workflows.