All case studiesProduction system
Invoice processing cut from three days to 15 minutes.
A beverage manufacturer matched supplier invoices against orders and receipts by hand, and processing took three days. A controlled three-way match now completes in 15 minutes, with 94% of invoices matched automatically.
- Client
- A beverage manufacturer
- Sector
- Manufacturing, finance operations
- Engagement
- Production system
- Measured result
- 3 days to 15 mininvoice processing
Open the workspaceSample data. No live systems connected.
INV-2051Vale IngredientsSample data
- Malt extract, 400 kg
- €6,120.00
- Hop pellets, 60 kg
- €1,980.00
- Freight
- €540.00
3 days to 15 mininvoice processing
Growth goal
Absorb a growing invoice volume without adding people to matching and receiving.
Operating constraint
Matching was manual, and warehouse receiving errors surfaced late as rework.
Measured result
Processing time down from three days to 15 minutes, with 94% of invoices matched automatically.
Summary
- Challenge
- Accounts payable compared each invoice with the purchase order and a paper receiving record. A receiving error sent the invoice back for another round.
- What SIEL built
- An AI model extracts supplier, order reference, lines, amounts and bank details. Deterministic rules match invoice, purchase order and goods receipt in the ERP on quantity and price.
- Controls
- No invoice reaches the payment run without a full match or a named approval. Every rule result and approval is logged.
- Impact
- Finance reviews exceptions with the evidence attached instead of matching invoices by hand.
The full account
01Volume growing faster than a manual match
The manufacturer's growth reached accounts payable as more invoices, more suppliers and more deliveries to check. Each invoice was compared by hand with its purchase order and with the warehouse receiving record, often on paper. Receiving errors surfaced only when the figures failed to agree, which meant repeated reconciliation. Handling more volume this way would have meant adding people to matching, which the finance team wanted to avoid.
02The model reads, rules decide, finance approves
An AI model reads incoming invoices, PDF or electronic, and extracts the fields the match needs. From there the decisions are deterministic. Rules run a three-way match of invoice, purchase order and goods receipt, within tolerances agreed with finance. A full match moves the invoice towards the payment run. Any difference goes to a named reviewer with the extracted lines shown next to the order and the receipt. The system prepares payments; a person releases them.
03What finance has now
Processing time fell from three days to 15 minutes, and 94% of invoices match without anyone handling them. The finance team works the remaining exceptions with the evidence in front of them, under its existing approval limits. The log shows, for every invoice, which rule passed or failed and who approved it, which is the record auditors ask for.
Discuss Your Operation With Our Engineers.
Describe one workflow and the systems it relies on. A senior engineer responds within two business days with an initial assessment: what we would build, what we would not, and why.