If you think about how most accounting workflows run today, they still follow a pretty familiar pattern. The bookkeeping gets done, the ledger is closed, and then someone senior steps in to review the file before anything moves into tax. This review step is crucial; it’s where errors are caught, balances are checked, and everything is brought into shape. However, it’s also where a lot of time disappears.
Even when the books appear finished, there’s usually another layer sitting underneath. Small inconsistencies, missing pieces, and transactions that don’t quite line up the way they should often remain unnoticed until the review stage. Having spent considerable time working in that space between bookkeeping and tax, I encountered this persistent problem: the middle step is still heavily manual.
The gap most firms work around
Over the years building LodgeIt, I observed this pattern repeating itself across various firms. Even when data flowed cleanly out of the ledger, it still needed substantial work before a tax return could be finalised. In practice, this gap typically manifests in several predictable ways:
• Transactions coded to the wrong accounts
• Hire purchase repayments missing interest components
• Journals needing adjustment before balances make sense
• Inconsistencies that only surface during review
None of this is unusual; it’s simply part of how accounting work gets done. However, it creates a dependency on senior reviewers stepping in at the end to fix what’s been overlooked. This reliance not only hampers efficiency but also increases the likelihood of errors slipping through the cracks.
What I set out to change
The concept behind the self-healing ledger is straightforward. Instead of relegating all that clean-up work to the review stage, we can bring it forward. The goal is to allow the system to address those issues earlier, directly within the ledger itself. By the time someone opens the file, they should not be starting from a list of problems; instead, they should start from something that already holds together.
This proactive approach has significant implications for both workflow and accuracy. It means less time spent on correcting errors and more time focusing on strategic insights and decision-making. With the self-healing ledger, the key is to empower the system to identify and rectify discrepancies before they escalate into larger issues.
What a self-healing ledger actually does
The easiest way to explain the functionality of a self-healing ledger is through an example. During a recent session, I walked through a typical scenario involving a hire purchase repayment that had been coded incorrectly. On top of that, the interest component was missing. This is a common issue, but here’s how the system handled it:
• It analysed the general ledger and ran validation checks.
• It identified the misclassification.
• It reallocated the transaction to the correct account.
• It inserted the missing interest component.
• It rebalanced the ledger.
• It confirmed that cash flow and roll forwards aligned correctly.
By the end of that process, everything reconciled cleanly, with no manual step-by-step review required to achieve this outcome. This capability significantly reduces the burden placed on accountants and allows them to focus on higher-value tasks.
Why this isn’t just another AI tool
Much of the current conversation around AI focuses on language models, which are useful for interpreting information, drafting content, and spotting patterns. However, accounting has a different set of requirements. The numbers need to reconcile, the logic must hold, and the output needs to be precise. This is where the current AI discussions often miss the point.
What’s truly needed is a combination of approaches working in unison:
• Deterministic rules to ensure calculations are correct
• Statistical and language models to interpret patterns
• Structured frameworks like XBRL to define how financial data behaves
The deterministic layer keeps everything reliable, while the AI layer assists the system in understanding the information it processes. This combination makes it possible to identify and fix issues without introducing uncertainty, thereby enhancing the accuracy and reliability of the accounting process.
Rethinking how financial data is structured
There’s another significant shift occurring beneath all of this. Most accounting outputs today are still built as documents—PDFs, spreadsheets, and reports designed primarily for human readability. This format inherently limits what systems can do.
When financial data is structured differently, the system can actually reason about it. It gains the ability to understand relationships such as:
• How profit flows into retained earnings
• How loan balances change over time
• How assets, liabilities, and equity maintain balance
Once these relationships are explicitly defined, the system can assess whether the ledger behaves as expected and can intervene when it doesn’t. This proactive capability is vital for ensuring accuracy and reliability within the accounting function.
Building a system that understands the ledger
The engine behind this innovative approach, which we refer to as JARVIS, is constructed around a knowledge model rather than a traditional database. Instead of treating entries as isolated data points, it works with the connections between them. This includes:
• Knowledge objects that capture relationships across the ledger
• Structured outputs that preserve the integrity of how data fits together
• The ability to verify accounting relationships across various reports
• Tax logic embedded alongside financial data
This system does more than simply process transactions; it evaluates how the entire ledger holds together. By integrating these advanced capabilities, we can ensure that the accounting process is both efficient and accurate.
What this changes for the review process
While the review stage isn’t going away, it will undergo a transformation. The essence of the review process remains, as it is still where judgement is applied. However, the starting point changes dramatically. Instead of opening a file and painstakingly working through issues one by one, you open a file that has already been tested and adjusted. This substantial shift allows professionals to focus on validating outcomes, reviewing exceptions, and applying judgement where it genuinely matters.
This represents a very different kind of work, one that can lead to increased job satisfaction and a more strategic role for accountants in their organisations.
Where this is heading
There is a broader shift underway in the accounting landscape. As systems improve at working with structured financial data, the way information moves starts to change. Instead of merely sharing documents, systems can exchange data in formats that other systems can directly understand. This advancement opens the door to more machine-to-machine interaction, allowing client systems and accounting systems to communicate seamlessly without needing everything translated into reports first.
This evolution will inevitably lead to a more integrated approach to accounting, where real-time data sharing becomes the norm, further enhancing efficiency and accuracy.
What this means for accountants
I do not foresee this shift removing the role of the accountant. Instead, it will change how and where your time is spent. There will be less time spent correcting the ledger and working through clean-up tasks. Instead, you will find yourself with more time to understand what the numbers mean and to help clients decide on their next steps.
The review process will still exist; it will just become more focused. Over time, this adjustment will shift the balance of work towards higher-value activities, ultimately enhancing the role of accountants as trusted advisors.
Watch the full session
If you want to see how I navigate these scenarios in practice, I cover it in detail during my QuickFest session. You can watch the full replay here:
