Manufacturing Firm Saves 20 Hours/Week with Power Automate

The Challenge

A manufacturing firm with 85 staff was processing purchase orders and supplier invoices across three separate systems that didn't talk to each other:

1. An older ERP system where POs were created
2. Email (Outlook) where supplier invoices arrived as PDFs
3. A spreadsheet maintained manually by the finance team to track approval status

The process looked like this:

• A staff member creates a PO in the ERP
• They email a copy to the relevant approver
• The approver replies by email to confirm
• Finance manually updates the spreadsheet
• When the invoice arrives from the supplier, someone manually matches it to the PO
• Another approval email goes out
• Finance updates the spreadsheet again
• Payment is processed

The whole process took an average of 3.5 days per PO. The firm processed 25–30 POs per week. Three different staff members each spent 5–7 hours per week on process administration.

Beyond the time cost, errors were common. Mismatched invoices, delayed approvals, and lost email threads caused supplier relationship issues and twice led to late payment penalties.

Our Approach

Phase 1: Process Mapping (Week 1)

We ran a two-hour workshop with the finance team, the operations manager, and a supplier-facing staff member. The goal was to map the current process in full — including all the unofficial workarounds that had accumulated over time.

The mapping session uncovered:

• Two additional approval paths (for POs above $10k and for capital expenditure) that weren't documented anywhere
• A workaround where some staff were bypassing the ERP and creating POs directly in spreadsheets "because it was faster"
• Different suppliers sending invoices in different formats (some PDF, some via a supplier portal, some via EDI)

Without this mapping exercise, the automation we built would have broken within weeks.

Phase 2: Tool Selection and Design (Week 2)

The firm was already on Microsoft 365. Power Automate was the obvious choice — it integrates natively with Outlook, SharePoint, and Teams, and the licencing was already in place.

We designed three connected flows:

Flow 1: PO Creation and Approval

• PO created in ERP triggers a SharePoint record
• Power Automate routes approval request to the correct approver via Teams (with email fallback)
• Approval or rejection is logged automatically
• ERP is updated via API

Flow 2: Invoice Processing

• New email with PDF attachment detected in the AP inbox
• AI Builder (part of Power Automate) extracts key fields from the invoice PDF
• System matches invoice to existing PO in SharePoint
• Discrepancies above threshold trigger manual review; matches route automatically to payment approval

Flow 3: Status Dashboard

• SharePoint list serves as the single source of truth for all PO and invoice status
• Automated daily summary to finance manager via Teams
• Overdue items escalated automatically after 48 hours

Phase 3: Build and Testing (Weeks 3–4)

We built the flows in a test environment using real (anonymised) data from the previous quarter. The finance team tested each scenario — including the edge cases uncovered in the mapping session.

Testing surfaced two issues:

1. Some supplier invoices had inconsistent field naming that confused the AI extraction
2. One approval path had a circular dependency (approver A needed to approve before approver B, but approver B's approval triggered a review by approver A)

Both were resolved before the flows went live.

Phase 4: Rollout and Training (Week 5–6)

Live rollout was a two-week parallel run — the old manual process ran alongside the automated flows so staff could validate the outputs. After two weeks with no discrepancies, the manual process was retired.

Training consisted of:

• A one-hour walkthrough session for the finance team
• Two short screen-recorded guides (how to handle exceptions, how to read the dashboard)
• A troubleshooting guide in the team's SharePoint

The Outcomes

Time saved: The three staff members who previously spent 5–7 hours each per week on process administration now spend less than one hour combined. That's over 20 hours per week returned to the business.

Error rate: Manual matching errors dropped by over 90%. The AI Builder extraction handles invoice data more accurately and consistently than manual entry.

Supplier relationships: Average PO processing time dropped from 3.5 days to 6 hours. Two suppliers who had raised concerns about payment delays both commented positively on the improvement within the first month.

ROI: The consulting engagement was recovered within 6 weeks of go-live, purely from staff time savings. The ongoing benefit continues at zero additional cost.

Key Lessons

Map before you automate. The two most critical things we discovered — the undocumented approval paths and the spreadsheet workaround — would have caused the automation to fail if we'd started building without the mapping session.

Start with the pain, not the technology. The business didn't need to know about Power Automate before this engagement. They had a pain (manual processing, errors, delays). The technology was chosen because it solved that pain efficiently, using tools they already owned.

Parallel running is worth the effort. The two-week parallel run felt slow, but it gave the finance team confidence and caught the edge cases that testing had missed. It's much easier to fix a flow before you've retired the manual process than after.