From 4 Hours to 18 Minutes: A Real Workflow Automation Case

When we started the engagement with this industrial manufacturer, their operations team was spending nearly half their working hours on work that shouldn’t require human judgment at all. Not complex decisions — data entry. Chasing emails. Copying numbers between systems. Running the same report manually every Monday morning.

230 hours a month. For a 12-person team, that’s a meaningful fraction of their capacity permanently allocated to being a human integration layer between software systems.

This is a writeup of what we actually built, how we approached it, and what surprised us.

The problem wasn’t automation — it was architecture

The team had tried to automate before. They’d built a few Zapier workflows that connected their email to a spreadsheet. They worked for a while, then broke when a supplier changed their email format. Nobody fixed them. Trust in automation evaporated.

When we came in, the first thing we did was map the actual data flows — not the official process diagram, but what was actually happening. What we found was that their production management system and SAP Business One were holding the same data, entered twice, by different people, with no reconciliation logic. Discrepancies were caught manually, days later, usually when something went wrong.

No automation layer built on top of that would have held. You’d just be automating the propagation of errors faster.

The decision we made before writing a single workflow: connect the systems properly first. Build a single source of truth for purchase orders, then automate on top of that foundation.

What we built

We used n8n as the workflow automation platform — self-hosted on their infrastructure, so all data stays inside their environment. Over three weeks, we built and validated 23 production workflows.

Purchase Order to Production Pipeline When a purchase order is approved, n8n pulls it from the procurement inbox, validates it against current production schedules and raw material availability in real time, pushes it to SAP B1, and sends a fulfilment notification to the plant floor team on Slack. End-to-end, under a minute. Previously: a manual process that averaged 4.2 hours.

Supplier Delivery Tracking Inbound supplier emails are parsed for shipment confirmations and ETA updates. When a confirmed delay exceeds a configurable threshold, an escalation fires to the relevant production planner — pre-populated with the affected PO number and the production runs it impacts. No manual monitoring, no delays in catching disruptions.

Material Shortage Alert Engine Raw material levels are checked every 15 minutes against live production demand. When a component crosses a reorder threshold, a Slack alert fires with a one-click draft purchase order already routed to the correct supplier. Previously, this required hourly manual checks of a dashboard that three people rotated through.

Automated Reporting Pipeline Data is pulled nightly from six source systems, standardised, and formatted into a weekly operations report that lands in twelve inboxes every Monday morning before the shift briefing. Previously: two full days across three people to assemble the same information.

The implementation reality

Three weeks, staged in parallel. We built each workflow in a test environment mirroring production, validated it against real historical data, and ran it alongside the manual process before cutting over. The operations team was involved throughout — not as observers but as collaborators who understood every decision point.

The things that took longer than expected: edge cases in supplier email parsing (every supplier writes differently), and a quirk in how SAP B1 handles certain PO status codes that wasn’t in any documentation. The things that went faster than expected: once the core integration was solid, layering additional workflows on top was straightforward.

The outcome

• Purchase order processing: 4.2 hours → 18 minutes
• Staff hours reclaimed monthly: 230
• Missed material shortage alerts in the months post go-live: zero (previously 3–5 per week)
• Annual labour cost recovered: $190,000
• Production disruption during implementation: none

The number that matters most to the operations manager isn’t any of those. It’s that her team now spends their time on production planning instead of data entry. That’s the actual value of workflow automation done properly.

If your operations team is spending significant time on manual data routing between systems, we should talk. We’ll map the workflows before recommending anything.