Intelligent Reconciliation: Match More, Review Less
Intercompany reconciliation is one of those processes where relatively small inconsistencies can create a surprising amount of work.
An invoice reference is entered differently by two entities. One side includes a prefix that the other doesn’t. Several payments settle a single invoice. Descriptions differ between systems, or the same transaction is described in different languages.
None of these necessarily represents a genuine reconciliation problem. An experienced accountant can often look at the transactions and see that they belong together. Traditional matching rules, however, tend to be less forgiving.
The result is that finance teams can spend a significant amount of time reviewing transactions that are not really exceptions. They’re simply valid matches that the software hasn’t been able to recognize.
Virtual Trader is changing that with Intelligent Reconciliation, an AI-assisted matching capability for Intercompany Cloud designed around a straightforward objective: match more transactions automatically, and give finance teams fewer exceptions to review.
Moving Beyond Exact-Match Reconciliation
Rules-based matching remains extremely useful. When transaction references, amounts, dates, currencies, and other relevant fields line up neatly, there is little reason to make the process more complicated.
Real-world intercompany data isn’t always that tidy, though.
This is particularly true in large multinational organizations. Transactions may originate in different ERP systems, follow different local processes, and contain references or descriptions entered by different teams. Even when two records clearly represent the same underlying transaction, differences in the data can prevent a conventional matching rule from bringing them together.
Intelligent Reconciliation adds another layer to this process by looking at the meaning and context of the data, rather than relying solely on exact text.
This means a mismatched reference number doesn’t necessarily result in a missed match. The same applies to formatting differences, minor errors, timing differences, and split settlements where a single invoice has been cleared by several payments.
The matching model is also multilingual. A description entered in one language can be compared with its counterpart in another, which is particularly useful when reconciling transactions across a global organization.
Fewer Transactions to Review, More Time for the Ones That Matter
The aim isn’t simply to achieve a higher automated match rate; it’s to change where finance teams spend their time.
If the system can identify the straightforward relationships with a high degree of confidence, there is little value in asking an accountant to confirm each one manually. Instead, reviewers can work from a confidence-ranked exception queue and concentrate on transactions where there is genuine uncertainty.
That has an obvious impact during the financial close. Reconciliation teams have fewer transactions to investigate, and genuine breaks become easier to identify among the noise.
It also makes the process more scalable. Intercompany transaction volumes tend to increase as organizations grow, acquire businesses, add entities, or introduce new ERP systems. Without greater automation, more transactions usually mean more reconciliation work. Intelligent matching helps break that cycle.
Perhaps more importantly, it makes better use of experienced finance people. Having an accountant investigate why two entities genuinely disagree is considerably more valuable than asking them to spot that INV-100234 and INV100234 probably refer to the same invoice.
AI Still Needs to Show Its Working
There is an important qualification to all of this.
For enterprise finance, getting the right answer isn’t enough; you also need to understand how the system arrived at it.
That becomes particularly important when AI is involved. A black box that produces impressive match rates but cannot explain its decisions is difficult to incorporate into a controlled financial process. Controllers need confidence in the result, reviewers need to understand what they are approving, and auditors may reasonably want to know how automated decisions were made.
So, every match proposed by Intelligent Reconciliation comes with a plain-English explanation of why it was made. It allows a reviewer to understand the relationship the model has identified, rather than simply being presented with a confidence score. Reviewer notes and a full audit trail provide a record of what happened.
Organizations can also configure tolerances, confidence thresholds, and the fields that matter for different data sets without changing code.
That gives finance teams control over how far they want automation to go. High-confidence matches can be approved straight through, while lower-confidence matches remain subject to human review.
We think that distinction matters. The purpose of AI in finance should be to reduce unnecessary human intervention without removing appropriate human oversight.
Keeping the AI Close to the Data
There is another aspect of Intelligent Reconciliation that we believe is particularly important for enterprise finance.
The matching capability is built on Oracle AI Vector Search, using a multilingual model that operates inside the Oracle database. Transaction data therefore doesn’t need to be sent to an external AI service for the matching process to operate.
This follows the approach we’ve taken more broadly with AI in Intercompany Cloud: wherever possible, intelligence should be embedded within the environment where the financial data and controls already reside.
For finance teams dealing with sensitive transaction data, that isn’t a minor architectural detail. It makes it easier to introduce more sophisticated AI capabilities without creating another external data flow that needs to be secured, governed, and audited.
From Reconciliation Processing to Exception Management
There has been a tendency to think about reconciliation automation in terms of making the existing process faster, whereas intelligent matching offers a slightly different opportunity.
If a system can identify a match confidently, explain why it believes the transactions belong together, and retain the evidence behind that decision, there is no particular benefit in putting every transaction in front of a person.
The role of the reconciliation team can increasingly become one of exception management rather than transaction matching.
That has implications beyond productivity. Genuine discrepancies become more visible because reviewers aren’t working through large volumes of false exceptions. Reconciliation becomes less of a month-end clean-up exercise, and finance teams have more time to investigate the underlying causes of the breaks that remain.
For large multinational organizations processing significant intercompany volumes, that is ultimately where we see the value of Intelligent Reconciliation.
Higher match rates are useful, of course. But the real outcome is simpler: match more of what doesn’t need human attention, so your people can spend their time on what does.
To learn more about Intelligent Reconciliation and how Virtual Trader’s Intercompany Cloud can automate reconciliation across complex, multi-ERP environments, get in touch with us.
