Enterprise project experience
Owners and managers use AI to see production, operations, and sales as a whole, connecting ERP, MES, CRM, WMS, OMS, TMS, inventory, and finance databases and APIs, with OpenClaw, Dify, n8n, and dedicated Agent Skill practice.
Scan a WeChat QR code on your phone to bind the AI assistant. No extra app or plugin. Owners can ask about the business by speaking, and stay on top of the full picture.
Scan in WeChat to bind. No extra plugin required.
Previously, getting a business answer required—
Exporting data, making reports, and repeated discussionsAsk the data directly
Waiting days for analysisGet actionable clues in minutes
Build a large system firstValidate value with a Skill first
Why not build another reporting system?
AI is good at reading data and finding patterns; what an enterprise needs to lock in is its own metrics, decision rules, actions, and safety boundaries.
A single executive question can send teams across departments to gather data, reconcile definitions, revise spreadsheets, and explain the results repeatedly. By the time the answer arrives days later, the business issue may already have changed.
Ask a question in WeChat and AI immediately queries authorized data. There is no need to write SQL or translate the request for technical staff first.
Production, order, and financial databases cannot be used for trial and error. A single write or delete command could disrupt live operations.
Use read-only accounts, query allowlists, least-privilege access, and audit logs. The Agent can view authorized data but cannot create, modify, or delete records.
After a report reaches a conclusion, teams still need another meeting to assign tasks and chase progress. Execution results never return to the same analytical workflow.
AI proposes an actionable plan that executives can forward to accountable owners in one step. After employees report progress in WeChat, AI automatically verifies whether the target metrics actually improved.
Repeated requirements discussions and ground-up development of interfaces and permissions drive up investment and delay launch. By the time the system is ready, the original problem may have changed.
Reuse existing AI and enterprise systems while codifying only the required metric definitions, business rules, and execution workflows. This is faster and more cost-effective than developing a new system from scratch.
AI Agent field practice
Connect ERP, MES, CRM, WMS, OMS, TMS, inventory, finance, and other systems through read-only databases and APIs. After binding in WeChat, leaders can question enterprise data directly and stay on top of production, operations, and sales.
Scan in WeChat to bind. No plugin to install. Owners can ask by voice or text, with no SQL.
Connect authorized data sources including ERP, MES, CRM, WMS, OMS, TMS, inventory, finance, reports, and files; database access is read-only by default, with clear and auditable permission boundaries.
Find anomalies and trace causes across dimensions, connecting data changes to business actions.
Bring together key changes across production, operations, and sales, then provide conclusions, evidence, and recommended next steps.
Technology selected to fit the need
General AI is good at fetching and analyzing data, but it does not natively understand each company's metrics, rules, and ways of working. General-purpose AI coding is still hard for non-programmers; I turn those specialist needs into repeatable Skills, faster and lighter than building a system from scratch.
For clearly scoped analytics and workflow needs, a Skill is usually leaner and faster than building a system from scratch.
Different industries, one closed loop
This goes beyond analyzing a report. AI must see what is happening, explain why, recommend action, and verify the outcome after employees report back. Each Agent can also be named to fit the company and use case.
Orders, production schedules, work in progress, and material arrivals are scattered across ERP, MES, and spreadsheets. Production managers spend two hours consolidating them every day, yet executives still see yesterday's picture.
“Which orders are likely to be delayed this week? Are materials, equipment, or staffing the constraint?”
AI cross-checks due dates, process progress, material shortages, and equipment downtime to identify high-risk orders, root causes, and affected customers.
Executives forward material-expediting and rescheduling lists to Procurement and the shop floor. After owners respond in WeChat, AI updates expected delivery dates and continues monitoring exceptions.
An order may be profitable when quoted, but exchange rates, ocean freight, material costs, and payment terms can all change before shipment. Sales teams focus on fulfillment, while Finance may not discover the loss until month-end.
“Which unshipped orders now have gross margins below 12%, and why?”
AI reads CRM, ERP, logistics, and financial data, recalculates current gross margin by order, and pinpoints changes in freight, exchange rates, or purchasing costs.
Executives assign freight renegotiation, quotation adjustments, and collections follow-up to business owners. Once results are reported in WeChat, AI automatically rechecks profitability.
Storefronts, advertising platforms, OMS, WMS, and Finance each report different figures. GMV may be rising while aggregate totals conceal high refunds, low margins, and SKUs approaching stockout.
“Which SKUs appeared to sell well yesterday but actually lost more money with every sale?”
AI combines advertising spend, platform commissions, refunds, purchase costs, and inventory to identify loss-making SKUs and distinguish acquisition, pricing, and product issues.
Operations receives a prioritized list of campaigns to pause, prices to adjust, and inventory to replenish. After changes are made, AI compares refund rates, gross margin, and inventory turnover to verify improvement.
Headquarters sees total inventory, while individual stores face either stockouts or overstock. By the time regional managers check every store through WeChat groups, the optimal transfer window has passed.
“Which fast-moving products will stock out within three days, and which nearby stores can supply them?”
AI combines store sales, on-hand inventory, inbound purchases, and delivery lead times to generate stockout risks and inter-store transfer recommendations.
Regional managers send transfer orders to store managers and warehouses. After receipt is confirmed, AI checks changes in stockout rates and slow-moving inventory.
Three additional capability areas
Each module has its own page with field scenes, cases, project outputs, and evidence.
Free personal project
A complete AI assistant for VS Code and Cursor, and public proof of Agent product and engineering capabilities.
View detailsBiomedical algorithm project experience
On high-dimensional, sparse, and complex relational data, I have practiced analysis, feature engineering, and deep learning.
View detailsSales and inventory forecasting project experience
Turn sales, pricing, promotion, and seasonality data into measurable forecasting models that support purchasing and business planning.
View detailsHow the project was done
Business goals, users, current workflows, and desired outcomes.
Confirm data quality, permission boundaries, and technical feasibility.
Validate Agent or model performance with a minimum viable version.
Improve from real feedback and write clear usage notes.