Challenge
Every audit engagement at this scale carries more than 2,000 individual IFRS, FRS, and GAAP checks, applied across every financial statement in the engagement. Standard OCR tools struggled with the complex, borderless tables common in financial reporting, which meant much of this checking still had to happen by hand.
The manual process also introduced inconsistency. Different reviewers applying checks to different sections of the same engagement could reach slightly different conclusions, not because the numbers disagreed, but because the reviewers did.
Solution
NexaQuanta built AI models that extract all tabular and narrative data from financial statements, understanding document structure and cross-section relationships well enough to auto-validate all 2,000-plus disclosure requirements per engagement. Rather than treating each table or paragraph in isolation, the system understands how sections of a financial statement relate to each other, which is what let it succeed where standard OCR had failed on complex, borderless tables.
The system runs on Azure Cloud with SUSE Linux and AKS for container orchestration, combining NLP, generative AI, embeddings, and PDF and Word document parsing in a single validation pipeline.
Results
Manual audit effort and time to completion: reduced by 80% per engagement.
● All 2,000-plus disclosure requirements are now validated automatically per engagement, regardless of table complexity.
● Consistent standards are applied across every section, removing the reviewer variance that came with manual cross-checking.
● Audit teams now spend their time on judgement-intensive review rather than repetitive validation.
What this means going forward
It took over the repetitive validation work so audit teams could spend their time on judgement instead of data entry. The AI validates to the standard the audit itself requires, which is why two of the Big Four trusted it with production engagements.
Ready to see what’s possible in your workflow?
Explore our Agentic Work Solutions →
