Bank Reconciliation Proof of Concept for Oracle
Middle East airline
In short
How can you test automated bank reconciliation before moving to a new ERP?
Matching bank transactions to finance records by hand costs time every day and ties the process to the people who know the rules. The client was moving to a new Oracle system and asked how reconciliation could work on it. We ran a short proof of concept on its own data.
At a glance
- Proof of concept in about three weeks
- Needed sample data and one 30-minute call
- People confirm every suggested match
The client
The client is an airline based in the Middle East, finishing a migration to a new Oracle system. It asked how automated bank reconciliation would work there.
What it was costing them
- Time: bank and ledger lines are matched one by one, every day.
- Key-person risk: matching rules live in people's heads, so the process depends on them.
- Growth ceiling: more transactions mean more people and more manual errors, not a faster process.
What we built
We ran a three-week proof of concept that showed a matching algorithm working on a sample of the client's own data.
- A light start: sample data, one 30-minute call and a workshop on the matching rules, with no system integration.
- Rules from the business: we captured the matching rules with the finance team, and added machine learning only where results can be checked.
- People stay in control: suggested matches are confirmed manually at first, then the team chooses which types clear automatically.
- Open items get owners: items with no match are assigned, the owner is notified and the reply is stored.
- Fits Oracle: a robot can work through the user interface, or use an API where Oracle exposes one.
- Better data in: extra identifier columns make matching more reliable.
Results
- The proof of concept ran for about three weeks after the sample data was handed over.
- Result: it showed a first match rate on real data, a base for a business case.
- Result: results were compared with today's manual matching, so any saving is shown, not assumed.
The risk most teams miss
The month-three risk is that a proof of concept looks good on a sample and then meets the real feed. We tested on post-migration data, kept human confirmation first, and agreed how match rate is measured.
What we would add today
Improvements we would add on top of what is live.
- Connect to Oracle through its REST API with a custom connector after the migration, keeping the robot as a fallback.
- Add a Dataverse model-driven app for review, with row-level security by bank account and legal entity.
- Read PDF bank statements with Azure Document Intelligence where no electronic feed exists.
Technology
Azure matching service, Power Apps, Dataverse, robotic process automation, Oracle ERP.
Tell us what your finance team matches by hand each day, and we will tell you what a three-week proof of concept would show.
Products and capabilities
- Power Apps
- bank reconciliation
- proof of concept
- matching algorithm
- Oracle integration
- robotic process automation
The service behind this work
Power Platform development
Related use cases
FinanceBank Reconciliation Automation with Azure, UiPath and Power Apps
Each day, the client's bank reconciliation team matched hundreds of transactions from two systems by hand, cleared them in SAP and chased owners by Outlook. The ERP's own matching rules were too limited, so the work, and the knowledge, sat with a few people. We built a daily cycle in which an Azure algorithm matches, a UiPath bot clears in SAP, and a Power Apps app keeps people in charge of every decision.
Schools Division Integrated Planning in the Accountability Framework App
Each year, the schools division network of the client had to plan business, manpower and finances for all of its schools and centres, in separate spreadsheets that someone had to merge by hand. The cycle had a fixed calendar and a presidential review, so late or inconsistent plans cost real time. We added integrated planning to its existing Accountability Framework app.
- 30+
- schools on shared templates
- c. 90
- departments imported
Weighbridge Integration and Automated Invoicing for Quarries
At a quarry every load is weighed, but the weight often stops at the gate. When tickets are keyed in again for invoicing, bills go out late, mistakes slip in and stock figures lag behind what has left the site. We linked the weighbridge to invoicing and to a live view of reserves.
Have a Microsoft project in mind?
Tell us what you need built, fixed or supported. A senior consultant replies within one business day.