Resources > Case Study > 1> Automating accounts payable invoice processing for Vistra with RPA and IDP
Automating accounts payable invoice processing for Vistra with RPA and IDP
AFusion automated Vistra’s accounts payable (AP) invoice processing with an RPA solution built on Intelligent Document Processing (IDP). The system extracts data from electronic invoices and scanned PDFs, classifies each invoice, matches it against related documents, handles discrepancies, and posts the result to the ERP — for both PO and non-PO invoices. After deployment, a single robot processes more than 2,000 invoices per month, replacing manual entry end to end.
Who Vistra is and what was breaking
Vistra is a global business services company operating in Vietnam, covering everything from company formation to post-setup support. Fast growth brought a fast-growing finance workload, and accounts payable felt it first. The AP team handled large volumes of invoices from many suppliers — electronic invoices and paper invoices side by side — under weekly payment runs and month-end reporting deadlines with very little slack.
Three problems came with that workload:
- Overload at predictable peaks. Invoice volume, supporting documents, and periodic financial reports all land in the same short windows. Staff stay overloaded, and the pressure falls hardest on the people closest to the deadlines.
- Repetitive, time-consuming tasks. Data entry, reconciliation, and transaction recording repeat every day. The repetition is boring, boring work breeds mistakes, and the mistakes cost real money in accounting.
- High error probability. Processing large amounts of data in short time windows is risky even for careful accountants. A small error in AP data distorts financial reports and can feed bad decisions upstream.
The cooperation goal
Vistra wanted one accounts payable automation solution that covered the whole chain: extract data from electronic and scanned/PDF invoices, classify each invoice, compare it with related documents, handle discrepancies, and enter the result into the ERP or accounting software. One hard requirement: the solution had to stay flexible across invoice types, handling both PO and non-PO cases.
How AFusion built the AP invoice automation
AFusion combined RPA with IDP to automate the accounts payable process. The robots are built with modern programming languages (C#, VB.NET, JavaScript), and large language models (LLM) support document understanding where templates vary. The workflow runs in four steps:
1. Extract invoice data with IDP
The IDP engine reads both electronic invoices and scanned invoices and pulls the fields that matter for AP: invoice code/number, issue date, supplier information, line items, quantity, unit price, totals, and tax. This removes the manual entry step that previously started every AP cycle.
2. Classify each invoice and apply the matching logic
The system sorts invoices into groups and applies the right checks to each. Purchase invoices are checked against the purchase order; service invoices are checked against contract terms. Classification is what lets one automation handle many invoice formats without a separate workflow for each.
3. Post to the accounting system through web automation
After classification and checks, RPA enters the invoice data into the accounting system through screen-level automation. Data reaches the ERP accurately and in near real time, without a human re-typing anything.
4. Notify the person in charge when something is wrong
Errors do not sit unnoticed. The system emails the responsible person as soon as it detects a problem — mismatched information, an invalid invoice, or a posting error — so exceptions get resolved the same day instead of at month-end.
Results after automation
- End-to-end automation. The full AP invoice processing flow — collecting, extracting, classifying, matching, posting to the ERP — now runs without manual entry.
- Productivity at scale. One robot handles more than 2,000 invoices per month, freeing staff for work that needs judgment.
- Accuracy by control, not by luck. Intelligent extraction plus strict validation rules reduce the error types that manual entry produces.
- Flexible and scalable. New processing logic can be added as processes or systems change, so the automation grows with the business.
- Maintained, not abandoned. AFusion runs a defined maintenance and monitoring process for the robots, keeping them stable after go-live.
Is your accounts payable process showing the same signs?
The pattern Vistra had is common across finance teams: growing invoice volume, weekly payment runs on short deadlines, month-end crunch, and staff whose day disappears into entry and checking. If that describes your team, the useful first step is not buying software — it is mapping where the invoices come from, what gets checked, and where the errors happen. AFusion works through exactly that mapping with finance teams, and shows which parts of an AP process automation can take over now and which need preparation first.
Frequently asked questions
How many invoices can one AP automation robot process?
In this project, one robot processed more than 2,000 invoices per month. Actual capacity depends on invoice formats, line-item complexity, and the validation rules configured for the ERP.
Can RPA handle non-PO invoices?
Yes. The solution classifies invoices by type and applies the matching logic per group — PO invoices are validated against the purchase order, while non-PO and service invoices follow their own checks, such as contract terms.
What happens when an invoice has an error?
The system flags the exception and emails the person in charge immediately, with the nature of the problem (mismatch, invalid invoice, posting error). Exceptions are reviewed and resolved individually instead of being discovered during reconciliation.
Does the solution write into the ERP?
Yes. After extraction and validation, the robot posts invoice data into the ERP/accounting software through screen-level automation, so the result is a real accounting entry, not just a data file.
Next Post
