You buy from 30–50 suppliers, place a few hundred POs a month, and the process looks like this: someone in purchasing opens Excel, checks stock in the ERP, retypes the lines into an email and sends it. Then waits for confirmation. Then types the confirmation back into the ERP. If that's your company — you've diagnosed the problem well. But the fix isn't as obvious as any vendor's "procurement workflow" deck makes it look.
Placing supplier POs is one of those activities that quietly balloons in small companies. Ten POs a week doesn't hurt. Fifty hurt. Two hundred hurt enough that a firm hires a second person just to keep the stream moving. And almost everything happening here is repetitive and mechanical. It's an ideal automation candidate — provided you know what you're automating, and why.
When it's even worth it
PO automation doesn't make sense for every company. It's worth it when at least two of these three apply: (1) you place more than 30 POs a week, (2) you have at least 15 active suppliers with repeat orders, (3) the process passes through more than one person (buyer places, manager approves, warehouse receives). Below that threshold, Excel plus a checklist in email is enough, and the automation will cost more to maintain than it saves.
The diagnostic we use up front: how many hours a week does your purchasing person spend on mechanical work — retyping lines, sending emails, chasing responses, entering confirmations back into the ERP? More than 8 hours — worth a look. More than 15 — you should tackle it this quarter.
Where the automation actually lives
Three realistic options, cheapest to most expensive.
- The purchasing module in your existing ERP. Comarch Optima, enova365, Subiekt GT, Symfonia — all of them have some sort of PO module. It usually can generate the document, save it to PDF, attach it to an email and send. Most often the best first investment, because the data lives where you need it — in the same ERP as your stock, supplier records and price lists.
- Middleware / a simple workflow (Power Automate, n8n, Make). Makes sense when the ERP's PO module is weak but the ERP does expose an API. Power Automate or n8n picks up the order record, generates a PDF from your template, sends the email with the purchasing@ mailbox in BCC, and logs the send back into the ERP. Cost: 200–800 PLN/month plus 2–4 weeks of implementation.
- A dedicated procurement portal. Coupa, SAP Ariba, smaller local platforms — makes sense when you run auctions, have many suppliers per category, and genuinely issue RFQs more than once a week. Below that — it's overkill.
Nine out of ten SMEs land on the first or second option. A procurement portal in a 20-person company is almost always a mistake.
What has to be in order first
Automation won't work on top of a database where one material has three names and "Chris from the metal shop" doesn't even have a tax ID in the supplier file. Before you send the first automated PO, sort out four things:
- The supplier file. One record per supplier, with the current tax ID, address, email of the person who takes orders, and (crucially) a flag for whether they accept PDF by email or require entry into their own portal. That one flag decides half the project.
- The item file. Your material code / SKU has to be consistent between your ERP and the supplier's catalogue. If it isn't — you need a mapping: "our XYZ-123 = their AB-4567". Boring, but without it the order flies blind.
- Prices and terms. Current supplier price lists have to sit in the ERP. Without them, the automation calculates the wrong order value, and the supplier either invoices their own way or (worse) ships less because "the order wasn't complete".
- Approval thresholds. Who signs off up to PLN 5k, who up to 20k, who above. That field belongs in the system, not in the owner's head.
What it looks like in numbers
At one of our clients, a 25-person upholstered furniture maker from the Wielkopolska region, the picture was this: 40 suppliers, roughly 400 PO lines a month, two people in purchasing. Each was spending about 15 hours a week issuing orders, sending them and recording confirmations. On top of that, 6–8 mistakes a month (wrong item, wrong quantity, wrong price), causing complaints back to suppliers and stoppages on the production floor.
Implementation took 6 weeks. We did three things: cleaned up the master data (suppliers + items), configured the purchasing module in their Comarch Optima to produce a branded PDF, and built a simple Power Automate workflow that emails the PDF to the supplier and tracks confirmation in the purchasing@ mailbox. Technical cost: about PLN 18,000 net plus PLN 400/month for the Power Automate licence.
After three months: PO time dropped from 30 to 8 hours per week (combined), order errors from 6–8 to 1 per month, and one of the two purchasers shifted more into supplier negotiation — which is what they were actually paid to do. Payback: around 7 months if you only count the recovered hours, distinctly faster if you also count the cost of production errors.
The traps we keep pulling clients out of
- "The supplier said they accept email" ≠ the supplier accepts email. In practice, roughly one in three Polish suppliers has a person who retypes emailed orders into SAP or Comarch. When that person is on holiday, your order doesn't exist. Automation doesn't fix this — it just surfaces the problem faster. Keep a fallback (phone, alternate contact).
- Automation without confirmation = shouting into the void. If your workflow just fires the PDF and marks the order placed — you have a problem. There must be a loop back: "supplier confirmed within 24 h" or "reminder" or "escalate to buyer".
- Approval thresholds in a comment on the email are not a control. If you want real control over spend, the workflow has to force a click on "approve" for anything over the threshold — not hope the manager reads the right sentence in the thread.
- Don't automate negotiation. Automated orders make sense for suppliers with a fixed price list. For materials whose price shifts every order (steel, timber, chemicals), automate sending the RFQ, but let a human still place the order itself.
Where to start
A realistic 2-week start:
- List the last 3 months of POs. Count: how many, to how many suppliers, how many lines.
- Pull the top 10 suppliers by PO count. They carry 70–80% of the volume — that's where automation pays off.
- Check whether your ERP has a purchasing module and what PDF it produces. It almost always exists — nobody has ever used it.
- Pilot with the 3 largest suppliers for a month: manual process change, PDF from ERP, email to supplier, confirmation logged. Prove the process before you add technology.
- Only then think about Power Automate / n8n / whatever. Automation on top of a mess produces a mess faster.
Total up-front investment: 2 weeks of clean-up + one week of pilot + 4–6 weeks of real rollout. No fantasy of "procurement transformation". Concrete effect — fewer hours, fewer errors, less stress on the shop floor.
If you'd like to talk through where your company is on this and what makes sense as a first step — the first call is free and takes half an hour.
Automation on top of a mess produces a mess faster. Prove the process works before you let technology near it.