Kim JunghoonKO

AI PO automation that keeps up with changing items and volumes

Period: 2026.06 – 10Stage: Development and integration testing

Project
PO automation & AX ERP development
Company
로보톰robotic furniture manufacturer
Scope
n8n documents and e-⁠mailERP approval and PDF tested
Automatic from shortfall calculation to per-supplier orderingBy hand: BOM review and requirement calculation, supplier details checked, then for each supplier a PO written, saved to Drive, and attached and sent by e-⁠mail, 3 steps repeated for 6 suppliers: 20 manual steps. With AI and n8n the manual work is 3 steps (production request entered, PO items approved, Slack run request); the AI shortfall calculation and the PDF and e-⁠mail per supplier run automaticallyRepeated per supplier · 3 steps × 6 suppliersManual handling20manual stepsAI and n8nautomation3manual stepsBOM review,requirementcalculationSupplierdetailscheckedPO writtenSaved to DriveE-⁠mail, attachand sendProductionrequestenteredAI shortfallcalculatedautomaticallyPO itemsapprovedSlack runrequestPDF and e-⁠mailper supplierDocuments and sending handled automatically per supplierManualAutomaticApprovalAutomatic from shortfall calculation to per-supplier orderingBy hand: BOM review and requirement calculation, supplier details checked, then for each supplier a PO written, saved to Drive, and attached and sent by e-⁠mail, 3 steps repeated for 6 suppliers: 20 manual steps. With AI and n8n the manual work is 3 steps (production request entered, PO items approved, Slack run request); the AI shortfall calculation and the PDF and e-⁠mail per supplier run automaticallyRepeated persupplier · 3 steps× 6 suppliersManual handling20manual stepsAI and n8n automation3manual stepsBOM review,requirementcalculationSupplierdetails checkedPO writtenSaved to DriveE-⁠mail, attachand sendProductionrequest enteredAI shortfallcalculatedautomaticallyPO itemsapprovedSlack run requestPDF and e-⁠mailper supplierDocuments and sendinghandled automaticallyper supplierManualAutomaticApproval
Automatic from shortfall calculation to per-supplier ordering · manual steps 20 → 3 ⁠·⁠ Diagram

AI shortfall calculation → Automated per-supplier ordering → AX ERP data integration

Key results and improvements

IndicatorResultBasis
ImmediacyShortfall calculation linked to orderingPO waits and repeated work kept to a minimumIn-house AI calculation → company form → sent to supplier
ScalabilityWide product mix and variable volumesLess manual work per itemShortfall calculation and per-supplier documents batched
ConsistencyPO documents and processing historySame criteria and form appliedCompany form, approval and sending history kept consistently

Challenge

  • As items and volumes grew, calculation, document preparation and sending were repeated over and over
  • Inventory, BOM, PO and approval history needed to be managed together

Work carried out

  1. AI and n8n PO logic

    Shortfall calculation, PDFs in the company's own form, e-⁠mail to each supplier

  2. React and Next.js web

    Screens for inventory lookup, PO drafts and approval results

  3. API and database design and integration

    A data structure for inventory, BOM, PO and approval history across 16 areas

  4. Function and integration tests

    ERP approval and PDF tests, and receipt of n8n PO e-⁠mails for 6 suppliers confirmed

Design decisions

  • AI computes changes in items and volumes; documents and sending are batched per supplier
  • PO approval and the execution history are checked by the person in charge
Scope of work
  • End to end on my own · automation, web, API and database design and development
  • Slack integration, approval, PDF and e-⁠mail tests, documentation
Coordination
  • Demonstrated the PO automation in in-house training
Applied skills
  • n8n
  • React
  • Next.js
  • TypeScript
  • PostgreSQL

AX ERP web build

Inventory lookup, PO approval and PDF generation built

React, Next.js · API, PostgreSQL | screens running in the development and test environment

Inventory and item lookup

Inventory overview and per-item lookup ⁠·⁠ Dev/test environment screen

PO approval and PDF

PO approval result, PDF and pending-send status · download of the generated PO PDF · receipt table left out of the image ⁠·⁠ Dev/test environment screen

ERP approval and PDF: tested in the development environment · n8n documents and e-⁠mail: tested separately · ERP e-⁠mail link: a later stage

  • Per-item inventory

    Item, quantity and status lookup · API and database link ⁠·⁠ Dev/test environment screen

  • PO approval result

    Approval outcome and pending-send status ⁠·⁠ Dev/test environment screen

PO output & receipt test

From a PO in the company form to the e-⁠mail received

Approved items are grouped by supplier and filled into the company form; PDF creation, saving and e-⁠mail sending run automatically

Generated PO · company form

PDF generated on 2026.06.03 · item table · test values and repeated rows included · raw test output ⁠·⁠ Test-generated document

Run request & e-⁠mail receipt test

Slack command부족수량에 대해 거래처 이메일에 발주진행해Place the orders with suppliers by e-⁠mail for the shortfall quantities

Slack run request → received in my own test inbox ⁠·⁠ Verification-run screen

  1. Production request and shortfall

    The AI PO system takes BOM, stock on hand and the production quantity

  2. Approval and supplier sorting

    Items approved by the person in charge are sorted by supplier and order method automatically

  3. Company form and PDF

    Items, quantities and unit prices filled into the form and saved as PDF

  4. E-⁠mail receipt test

    Documents and e-⁠mails for 6 suppliers confirmed · my own test inbox used

The PO PDF and the inbox screen are separate test records of the same automation, distinct from actual receipt by suppliers or ongoing operation. Manual steps 20 → 3 compares work steps for 6 suppliers.

Expansion structure

AI PO calculation and the manufacturing-operations data structure

Shortfall calculation from production requests, inventory and BOM, with PO and approval history in one structure

AI PO calculation and the manufacturing-operations data structureA request from the production request screen goes to the server, whose AI shortfall calculation records it in the inventory and PO database; the draft passes approval by the person in charge and becomes a PO PDF filled automatically in the company form; e-⁠mail sending through the ERP is later integration scope. The database covers 16 business areas such as inventory and BOM, purchase requests and POs, approval and change history, and warehouse, production and quality. The AI agent is read-only and answers from database records through a separate read permission. Equipment, WMS and MES are the expansion directionOrderingExpansion directionProductionrequest screenProduction quantityEntry and runAI agentAnswers fromdatabase recordsRead-onlyServerAI shortfallcalculationSeparate readpermissionInventory andPO databaseInventory · BOMPurchaserequests · POsApproval ·change historyWarehouse,production,quality16 business areasApprovalRequesterand approverkept separatePO PDFCompany formfilled automaticallyE-⁠mail sendingERP sending linkLater integrationscoperequestrecorddraftapproveattachaskbasisreadEquipmentPLC dataWMSInbound, storage,outboundMESWork orders, resultsWeb and database builtApproval by the person in chargeExpansion directionAI PO calculation and the manufacturing-operations data structureA request from the production request screen goes to the server, whose AI shortfall calculation records it in the inventory and PO database; the draft passes approval by the person in charge and becomes a PO PDF filled automatically in the company form; e-⁠mail sending through the ERP is later integration scope. The database covers 16 business areas such as inventory and BOM, purchase requests and POs, approval and change history, and warehouse, production and quality. The AI agent is read-only and answers from database records through a separate read permission. Equipment, WMS and MES are the expansion directionOrderingProductionrequest screenProduction quantityEntry and runrequestServerAI shortfallcalculationSeparate readpermissionAI agentAnswers fromdatabase recordsRead-onlyaskbasisrecordreadInventory andPO databaseInventory · BOMPurchaserequests · POsApproval ·change historyWarehouse,production, quality16 business areasExpansion directionEquipmentPLC dataWMSInbound,storage,outboundMESWork orders,resultsdraftApprovalRequester and approverkept separateapprovePO PDFCompany formfilled automaticallyattachE-⁠mail sendingERP sending linkLater integration scopeWeb and database builtApproval by the person in chargeExpansion direction
Solid: web and database built · grey: approval by the person in charge and later integration · dashed: expansion direction ⁠·⁠ Diagram

Built, verified and expansion scope

IndicatorResultBasis
BuiltInventory and PO systemWeb, API, databaseBOM of 85 partsapproval and change history
VerifiedDevelopment and test environmentApproval, PDF, e-⁠mailERP approval and PDF / n8n PO e-⁠mail
Expansion designConnecting manufacturing systemsEquipment, WMS, MES98-⁠page system design standard

Test scopen8n: PO and e-⁠mail receipt tested for 6 suppliers. ERP: approval and PDF functions tested. The ERP e-⁠mail sending link is later scope.

Disclosure scope Dev/test environment screen 4 · Verification-run screen 1 · Test-generated document 1 · Diagram 2. The ERP screens run in the development and test environment, the PO is the item table of a PDF generated in an n8n test, and the e-⁠mail is from my own test inbox. Only part of each screen is shown here.

For more material, please ask by e-⁠mail.

fpdlsdk3115@gmail.com

 ⁠·⁠