Automating EDI with n8n: from duplicated email to the ERP in a single step

The order that got duplicated three times

In international B2B retail, the most expensive problem isn't missing technology: it's the manual translation between formats. The customer sends the order by email in free format, a human types it into the ERP, the system rejects it, another human types it again — and the SLA drops while the customer waits.

This pattern cost me two years of real operations. The measurable impact once we automated it: 40% improvement in the document workflow SLA. Not by building new software, but by replacing the human translation layer with an n8n workflow.

Why EDI is still alive (and still hurting)

EDI exists because large retailers (Whole Foods, Walmart, Safeway) require it as an operational contract. But EDI doesn't talk to your ERP out of the box. There's a gap in the middle:

• The EDI 850 (purchase order) the retailer sends.
• Your ERP that expects a different format.
• Validation rules (SKU, quantity, address, lead time) that are yours.

In many teams a human fills that gap. That's exactly where n8n fits.

The topology I used

EDI 850 (email / VAN) → n8n intake → normalize to JSON
  → validate SKU + quantity + lead time
  → if passes: create order in ERP
  → if fails: separate ticket to a human with full context

The key decision is what goes to a human and what goes to the system. Whoever automates 100% doesn't scale; the 5% that breaks the rule accumulates and erodes the retailer relationship.

What I learned:

• A human with full context is cheaper than a human without context. Before, the operator opened the email, opened the ERP, opened the ticket system. Now, every exception arrives with the already-normalized JSON, the rules that failed clearly flagged, and the ticket draft prepared.
• Measure exception rate, not just throughput. If the flow breaks 20% of the time, you'll drift without knowing because the overall SLA still looks fine: exceptions are absorbed by invisible overwork.
• The customer doesn't need to understand EDI to see the result. The indicator that connects most is "your last order was processed in X minutes, untouched by hand". That's more concrete than any dashboard.

The commercial decision almost nobody makes

The real bottleneck of EDI automation isn't technical. It's the decision of where to cut automation and leave an exception lane for humans. My rule:

1. If the failure is format or unknown SKU → escalate to a human, with the ticket draft already prepared.
2. If the failure is address or lead time → block the ERP send and notify the customer on their preferred channel with the exact detail.
3. If everything validates → it goes straight to the ERP and the customer gets immediate confirmation.

Outcomes you can feel: the customer stops asking "did my order arrive?". In many teams a CSM spends 30–50% of their time answering that question. If the system eliminates it, you free CSM capacity for conversations that actually move retention.

How to start in a week

1. Pick the 3 order types that move the most volume and SLA. Don't try everything at once.
2. Stand up the intake webhook (n8n accepts email-in, public webhook or SFTP reads). The channel depends on the retailer.
3. Build the SKU + quantity validator only. It's the piece that unlocks the most value.
4. Measure the clean-exception rate (with context). If it's > 15%, your validation rules are off; worth revising them before moving forward.

That's everything you need for the first experiment. If you want it in your operation with your real EDI and your real ERP, I respond within 1 business day with a scoping brief that maps your channel → your ERP → your volume.