The everyday jobs are one-shot requests. Workflows are the chains that run themselves: a shipment finishes, an invoice follows, the customer gets mail, the EDI file lands in the mailbox. Your Vibe-code agent can build these as scripts and cron jobs on the server, wired into metasfresh’s database and REST API. Here are ten worth setting up.
1. Order-to-Cash Without the Copy-Paste
Completed shipment, generated invoice, emailed PDF, despatch advice to the customer’s EDI mailbox — four steps your back office does by hand for every order. The agent can chain them into one scheduled job: find shipments completed that day, run the invoice candidates into real invoices, email the PDFs to each customer’s order contact, and write the DESADV file for the EDI folder. You review a summary instead of clicking through the billing window. Try: “Every evening at 6 PM, find shipments completed that day, generate the invoices, email them to each customer’s order address, and drop a DESADV file in the EDI outbox.”
2. Sync Stock and Prices to Your Storefront
If you sell online alongside your wholesale trade, the webshop must reflect what metasfresh actually has. A nightly job pulls on-hand quantities and the B2C price list through the REST API, pushes them to WooCommerce, and flags anything that failed to sync. Overselling drops, and you stop maintaining two price lists by hand. Try: “Write a nightly job that pushes current on-hand quantities and the B2C price list prices from metasfresh to our WooCommerce shop, and alert me in chat if any product fails to sync.”
3. Manufacturing Triggered by Demand
When a big sales order reserves more of a manufactured product than you have, someone has to run planning, create the manufacturing order for the finished good, and raise purchase candidates for any BOM component that is short. That is a fixed chain the agent can automate: check reservations, create the manufacturing order with the right bill of materials, and push missing components into purchase disposition with the needed date. Try: “When a sales order reserves more than we have on hand for a manufactured product, create its manufacturing order from the BOM and raise purchase candidates for any component we are short of.”
4. Contract Renewals on Autopilot
Delivery and subscription contracts in metasfresh renew automatically, but the commercial side — confirming the renewal with the customer at current prices — is usually manual. A weekly job lists contracts expiring in the next 30 days, prepares the renewal sales orders at each customer’s current price list version, and sends you the list to approve before anything goes out. Nothing slips through the cracks in a busy month. Try: “Every Monday, list contracts expiring in the next 30 days, prepare their renewal sales orders at current prices, and email me the list for approval before anything is sent.”
5. From RFQ to Best Offer
For bigger purchase lines it pays to ask several vendors. The agent takes the purchase candidates over your threshold, drafts requests for proposal to your registered suppliers, and when the quotes come back as CSV or PDF it compares price, delivery date and minimum quantity, then creates the purchase order for the best one that meets the date. Try: “For purchase candidates over 5,000 EUR, send RFQs to our three registered suppliers for that product category, and when quotes come back, compare them and create the PO for the cheapest one that can deliver by the needed date.”
6. Payment Reconciliation That Matches Itself
Incoming SEPA payments arrive with references that almost match your invoices. The agent imports the bank statement, matches payments to open invoices by amount and reference, clears differences under a threshold to a rounding account, and leaves a short list of unmatched items for a human. Matched invoices get paid, which also releases any dunning or credit hold. Try: “Import last week’s bank statement CSV, match incoming payments to open invoices by amount and reference, clear differences under 1 EUR to the rounding account, and list anything unmatched for me.”
7. The Dunning Ladder, Weekly and Calm
Dunning works best when it runs on a fixed rhythm instead of when someone remembers. A weekly cron job creates dunning candidates, emails level-1 and level-2 notices with your standard texts, adds level-3 accounts to a collections sheet, and posts a summary of who moved up a level. Chronic cases surface on their own. Try: “Set up a Friday 5 PM job that creates dunning candidates, emails notices for levels 1 and 2, adds level-3 accounts to the collections list, and posts a summary in chat.”
8. Shipping-Day Pipeline for the Warehouse
On shipping days the warehouse needs a list: what is due, in what order, grouped so picking walks are short. The agent picks all sales orders due today, groups them into picking tours by warehouse zone, prints the picking lists, and after each shipment is confirmed sends the customer a short “your order has shipped” mail. The packers stop asking the office what goes out. Try: “At 7 AM on shipping days, collect all sales orders due today, group them into picking tours by zone, print picking lists, and send each customer a shipped confirmation when their shipment is completed.”
9. Quality Gate at Goods Receipt
For sensitive goods — food, chemicals, anything with a shelf life — the receipt is the moment to catch a bad lot. The agent can enforce it in the receipt flow: lot number, best-before date and a photo are required before the receipt can be completed, and if a measured attribute like temperature is out of range the batch is blocked and the quality contact is notified. Try: “For receipts from supplier FrischeParadies, require lot number, best-before date and a photo before the receipt can be completed, and if the temperature attribute is out of range, block the batch and notify the QC user.”
10. Supplier Confirmations Back Into the System
Suppliers confirm your purchase orders by email with a promised date that rarely matches what you asked for. The agent watches the purchasing inbox, pulls the promised date out of each confirmation, updates the purchase order, and flags any that will arrive late so you can replan the sales orders that depend on them. Try: “Watch the purchasing inbox for order confirmations, update each purchase order’s promised delivery date from the PDF, and list any that will arrive after the date we requested.”
Each of these is a script plus a schedule, so you can inspect exactly what runs on your server. Start with the one that eats the most hours, and ask the agent to dry-run it before it goes live.
See metasfresh on OpenSysLab for the server that runs all of this out of the box.