How a Construction Company Eliminated Improper Payments Without Growing Its Finance Team
Introduction
A construction company with more than 200 active suppliers faced a problem that only surfaced after the payment had already gone out. Purchase orders, invoices, and payment slips did not always match, and manual checking covered only a fraction of the volume.
The problem
Every supplier delivery generates three documents that need to match each other: the purchase order, the invoice (nota fiscal, Brazil's mandatory tax invoice), and the payment slip (boleto, Brazil's standard bank payment document). Tax ID (CNPJ), amount, and document number need to be identical across all three before any payment gets approved. With 200 active suppliers, the volume of checks grows faster than the finance team's manual review capacity can keep up.
In practice, the team reviewed what fit within the approval window and moved on. Whatever fell outside that review became a silent exception: a mismatched tax ID, a different amount on the payment slip, a document with no match. These cases only surfaced later, once the payment had already been made.
In an operation with hundreds of suppliers, it took only a few occurrences a month to generate rework, strained supplier relationships, and loss of control over the process. The result was recurring: between 3 and 5 improper payments per month, caught too late to reverse smoothly.
The problem was one of operational capacity. Manually reviewing 13 fields across three documents, supplier by supplier, took 15 to 20 minutes per check.
At volume, available time always falls behind what's needed.
What got checked on every payment
Each check cross-referenced fields such as:
- Supplier tax ID (CNPJ) across all three documents
- Invoice amount
- Payment slip amount
- Purchase order number
- Invoice number
- Issue date
Thirteen fields in total, repeated for every delivery from each of the 200+ suppliers. That volume, multiplied by manual review time, is what creates the bottleneck.
How manual supplier verification used to work
The check remained 100% manual. The finance team opened all three documents side by side, compared them field by field, and approved the payment when everything matched. This process covered the fraction of volume that fit within the team's available time. The rest depended on whatever capacity the team had available at that moment.
That gap had a measurable cost: recurring improper payments, uncertain traceability over which step should have caught the divergence, and a finance team tied up in repetitive checking instead of analysis.
The improper payment was only the symptom
When improper payments show up, the most common reaction is to put more pressure on the finance team or add new approval steps. One more approver, one more check, more pressure to review carefully.
These measures raise the pressure on the team but leave the bottleneck intact: the time it takes to check each document stays the same. The real problem was the capacity to review 100% of documents within the window required for approval.
As the supplier base grew, checking capacity stayed the same size, limited to the team and the time each analyst had available. At some point, the operation starts working by sampling, even when no one made that decision formally. That's the point where divergences stop being isolated exceptions and become a natural consequence of the process.
How to automate the match between purchase order, invoice, and payment slip
JedAI started checking the purchase order, invoice, and payment slip the moment all three documents arrive, before payment approval. The 13 critical fields are cross-checked automatically across the documents, with any divergence flagged before the money goes out.
The logic is simple to describe and hard to execute manually at scale. Purchase order matches, invoice matches, payment slip with a divergent amount blocks the payment until an analyst handles the exception. The finance team now acts only on the exceptions JedAI flags, exactly where human judgment is needed, instead of reviewing document by document.
Results from automating document verification
- From 15-20 minutes to under 15 seconds per check across the three documents
- 100% of documents reviewed, covering the full monthly volume
- Zero improper payments after implementation
- Automatic detection of tax ID, amount, and document number divergences before payment approval
- Finance team reallocated from repetitive checking to exception analysis
The real difference is in coverage. The company went from reviewing a fraction of volume to reviewing 100%, with the same team. The bottleneck was structural: the process depended on human capacity to scale.
What this case means for companies with a high volume of suppliers
This math applies to any operation with a high volume of suppliers and manual document verification. Construction, but also retail, manufacturing, and distribution face the same pattern: volume grows faster than manual review capacity, and the gap turns into an improper payment or a divergence discovered too late.
The question worth asking is simple: does verification cover the full volume, or only the fraction that fits in the available time? The difference between the two only shows up after the payment goes out.
Automating document verification redirects the finance team's judgment. It removes the repetitive task of comparing field by field, document by document, and gives back the capacity to act where the decision matters: on the exceptions.
Does your operation review 100% of documents, or work by sampling?
If the answer is sampling, it's worth understanding where the bottleneck is. Talk to a CADS specialist and see how to automate the match between purchase orders, invoices, and payment slips without growing your finance team.