By: Aiki Kuldkepp
05 Nov 2026 8 min read
We are happy to provide you with new (international) insights in the areas of finance, business operations, strategy, governance, risk, compliance and more.
As a result, an audit may begin with questions about thousands of transactions that share the same discrepancy, rather than a request for a sample of records. Businesses will increasingly need to explain not only individual invoices, but also the VAT decisions, controls and processes behind them. What does this mean in practice?
E-invoicing makes invoice information machine-readable. Digital reporting gives tax authorities earlier access to transaction data that falls within reporting obligations.
Under ViDA, Digital Reporting Requirements for certain cross-border B2B transactions will apply from 1 July 2030. However, ViDA is not the only driver behind this development. National e-invoicing and digital reporting regimes, together with tax administrations’ growing use of data analytics, are contributing to a broader shift towards digital, data-driven VAT compliance and enforcement.
Auditors traditionally use sampling because reviewing every invoice individually is time-consuming. E-invoicing and digital reporting make structured data readily available, creating new opportunities for auditing. Software can review all available records against predefined criteria and identify recurring discrepancies.
The scope of these reviews will depend on national reporting requirements, access to data and the analytical capabilities available to tax authorities. However, reviewing more transactions does not automatically provide all the information needed to assess the VAT treatment of those transactions. Tax authorities may raise questions based on digitally reported data, but businesses must still maintain the records required to support and justify the VAT treatment applied.
Not every automated check requires AI. Duplicate invoices, calculation errors and discrepancies against VAT rate tables can already be identified through predefined rules. Machine learning, however, can help detect less obvious patterns and identify transactions that warrant further investigation.
OECD reporting already describes tax administrations using AI for anomaly detection and risk assessment. This is no longer a purely theoretical development. Tax administrations are increasingly exploring and implementing AI and advanced data analytics as part of their compliance and risk management activities.
For VAT purposes, this may include identifying apparently similar transactions that receive different VAT treatments, unusual combinations of transaction characteristics or recurring discrepancies that merit further review. Nevertheless, those differences may be fully justified by the underlying contract or supply chain. AI can help identify the question, but the answer still requires a careful assessment of the facts and the applicable VAT rules.
Digital reporting means transaction data can increasingly be reviewed alongside related records rather than in isolation. Where information from both parties is available, tax authorities may compare the data reported by the seller with the data reported by the buyer. As a result, digital reporting of cross-border transactions creates additional opportunities to identify discrepancies and potential fraud. The focus shifts from whether an invoice appears correct on its own to whether the information reported by all parties is consistent.
The same principle applies to credit notes and, where available, customs information. Consider a distributor that grants retrospective volume rebates. Repeated credit notes that do not include references linking them to the original invoices may trigger requests for clarification. This does not mean the VAT treatment is incorrect. However, it does mean the business should be able to reconcile the reductions with the underlying sales and explain how they were reported. The audit question becomes more specific: how does the original transaction connect to its subsequent correction?
An error on a single invoice may involve only a limited amount of VAT. However, if the same tax code is applied incorrectly across thousands of invoices, the underlying issue becomes significantly more important. Reviewing the full transaction population helps distinguish a one-off discrepancy from a recurring system error.
For businesses, correcting an affected VAT return is only part of the solution. They must also determine what caused the error, identify which earlier transactions were affected and assess whether the underlying process continues to generate the same issue. By contrast, a documented timing difference may simply require an explanation rather than a system change.
Successfully sending and receiving an e-invoice, and transmitting transaction data to the tax authorities, confirms only that the data has been exchanged. It does not confirm that the VAT treatment is correct.
Businesses should therefore test the information they intend to report, reconcile corrections and investigate discrepancies they cannot explain. There is no need to replicate a tax authority’s confidential risk model to do this. Practical checks can already uncover recurring issues.
As tax authorities expand their use of data analytics and AI, businesses have a strong incentive to strengthen their own analytical capabilities. This can help identify recurring discrepancies and inconsistent VAT treatments before they become subjects of a tax audit.
At the same time, it is important to distinguish between errors that require correction and unusual transactions that require specialist review. Blocking every exception may disrupt legitimate business activities, while ignoring every warning without investigation leaves underlying risks unresolved.
An invoice shows how a transaction was treated for VAT purposes, but it rarely explains the reasoning behind that treatment. Supporting the VAT position may still require evidence such as the VAT status of the customer, contractual documentation, details of where goods were shipped from and to, transport evidence, and information about the timing or nature of the goods or services supplied.
These records should be linked to both the transaction and the reported data. Businesses also need to know which customer information, tax rules and system settings were in place when the invoice was issued. Current settings may not explain an invoice issued in an earlier period.
Preparing for a more data-driven audit means being able to reconstruct the original VAT decision and any subsequent correction without having to assemble the evidence only after questions have been raised.
Not necessarily. Automated analysis can cover a larger volume of transactions, but available data may still be incomplete or reported at different times. A discrepancy alone is not evidence of an error. Likewise, invoice data does not prove that goods were transported to another EU country or that all substantive conditions for a VAT exemption were satisfied.
Tax authorities will still need to review supporting evidence and allow businesses to explain apparent inconsistencies. Secure data handling and proportionate investigations will remain essential. Sampling may also continue to play a role in testing evidence or identifying risks that analytical models fail to detect.
As a result, broader automated screening is likely to complement rather than replace traditional sampling and other established audit techniques.
Preparation for digital reporting should extend beyond converting invoices into the required format. Businesses need to understand whether their transaction data is consistent, whether their VAT determination logic functions correctly, whether the correct tax codes are being applied, whether corrections can be traced and whether they can support the VAT treatment adopted.
An accurate aggregate VAT return remains important, but it will not answer every question generated by transaction-level analysis. By carrying out their own checks, businesses can identify recurring errors before a tax authority does. This becomes increasingly important as tax administrations make greater use of data analytics and AI to identify risks and anomalies across large volumes of transaction data.
The practical starting point is the data that businesses already have available: test it, identify patterns, investigate unexplained discrepancies and address their root causes. This makes e-invoicing preparation valuable for VAT compliance well before an audit begins and helps businesses prepare for an increasingly data-driven approach to VAT enforcement.
Would you like to discuss these insights? Get in touch with us.
For many businesses, VAT has traditionally been treated as a reporting exercise: transactions take place, invoices are issued and the tax team reviews the information afterwards when preparing the VAT return.
The Netherlands has reached an important milestone in implementing the European Union’s VAT in the Digital Age (ViDA) package. Following adoption by the House of Representatives on 17 September 2026 and by the Senate on 6 October 2026, Bill 36 920 has now been approved by both Houses of Parliament. The bill implements the first stage of ViDA in the Netherlands: the Single VAT Registration (SVR) measures.
The Netherlands has opted for a broad implementation of ViDA (VAT in the Digital Age) by introducing mandatory e-invoicing and digital VAT reporting not only for cross-border transactions, but also for domestic B2B transactions.