Manually reconciling accounts or relying on an increasingly complex set of macros in Excel consumes precious time better spent on something else. But investing in new technologies or process improvements present their own risks: Will it work? What will it really cost us? Is it going to break something? This article outlines five types of risk you can reduce through by moving to a platform built on Machine Learning (a form of Artificial Intelligence) and sophisticated software, while also helping your team move beyond manual reconciliations and achieving a more efficient close process.

Five Types of Risk Reduced:

• Internal Control Risks
• Human Error Risks
• Risk of Increased Work Loads
• Audit & Restatement Risk
• Purchase Risk

Internal Control risk

‍In a manual or Excel-driven reconciliation world, COSO 13 controls are fragile at best and non-existent at worst. This Internal Control risk leaves F&A teams fully exposed across a wide swath of F&A fiduciary responsibilities - from accuracy of financial statements, to protection of assets, to fraud prevention..

A machine learning + centralized software approach to account reconciliation reduces Internal Control risk by:

• Controlling data access. Anyone with access to folders in your storage directory also has access to your Excel spreadsheets. Even with password protected spreadsheets, passwords are distributed and shared all along the workflow process. A centralized account reconciliation platform has built-in access controls, advanced security, and segregated roles that define access - for example, admin, staff and approver. User access activities are recorded and logged for review or audit. Excel-based reconciliations simply cannot do this.
• Protecting data from accidental or deliberate changes. Modifying data is frowned upon by auditors. In an Excel-based reconciliation process, there are no records of who made a change or what was the change. Along with access controls, COSO 13 compliant software platforms also prevent data modifications while recording and logging changes to derived data within the user’s control (for example, un-matching and auto-matched transaction or designating a manual match).
• Preventing formula mistakes, overwrites, and macros corruption. Any person in the F&A function for any length of time has experienced the corruption of a macro, accidental overwrite of an Excel formula, or simply crafting the wrong formula. Often these go undetected for months, resulting in publishing wrong numbers, internal restatements, or worse - external restatements. The algorithms developed by a machine learning approach cannot be changed by the end user and become better as they are applied to many different types of reconciliations unlike fragile Excel formulas.
• Reviews and approvals are easily tracked. A centralized software based system for transaction matching will serve as a source of record for reconciliations including who completed the tasks and who approved them.

Human Error Risks

‍The human mind is an amazing computational organ. However, we are limited by how much data we can process visually – scanning sorted rows of transactions for example – and are prone to mis-read errors and the bias to be finished with our work. Whether keying in data incorrectly, inadvertently changing data format or row header in Excel, or failing to set-up a process when you go on vacation, humans are wonderful but not perfect.

A machine learning + software approach changes the responsibility of staff from one of a transaction matcher, to more of a reviewer and researcher. Modern machine learning based systems often record matching performance of 99%+ which means you can focus more on issues that need to be resolved vs manually managing the transaction matching process.

Risk of Increased Work Loads

F&A teams often resort to batch-level reconciliations when they should truly match transactions, simply because of the sheer time needed to manually match. Yet, how many times have you run a batch level reconciliation only to find that the adjusted-totals don’t agree? The attempt to avoid time spent manually matching is quickly squandered as you go “treasure hunting” to find reconciling items that throw off your batch totals, and you end of spending far more time than if you had bitten the bullet and manually matched from the outset.

A modern machine learning + software approach can easily move you from a batch process to a transaction-level matching process because it can efficiently handle thousands or millions of transactions with a high degree of match accuracy. A system like Sigma IQ also identifies individual transactions that need your attention vs you having to spend hours (or days) finding them.

Audit and Restatement Risk

No one likes an qualified audit opinion, a material weakness finding, or having to do a restatement because of either poor process or internal controls (see risk one). It’s also not good for your career. A machine learning + software approach to account reconciliation can minimize audit and restatement risk in two ways:

• 100% auditability. Every activity and piece of data that is processed through a modern, cloud based platform should be captured at the user level for audit and review purposes. If built properly, these systems - like Sigma IQ – can even re-run or re-play every reconciliation as desired in a fraction of the time it takes to re-run manually.
• Increased accuracy. Even best-of-class rules-based reconciliation platforms commonly achieve 60-75% automated matching on a regular basis, leaving the remainder for manual matching and increasing the risk of errors. Manual reconciliation, when performed at scale, is worse. A well functioning machine learning based approach should be capable of achieving 99% matching performance across dozens of different reconciliation use cases.

Purchase Risk

‍There is always risk in bringing on a new technology or new process. Will it work as expected? Will it cost more to implement – in time or dollars – than promised? Will it really reduce my workload and provide an ROI?

Traditional account reconciliation software solutions are based on creating a series of specific rules to ingest and match transactions. This takes time, budget, and specific skills to implement – often by a system integrator team. You won’t know until after the system is implemented whether or how beneficial it is compared to your current process.

A machine learning approach to reconciliation can reduce purchase risk in four specific ways:

• Will it work with my data? A proof-of-concept (POC) using your actual data provides you the best information whether the system will work for your data. Machine learning solutions should be able to easily learn from and process your data so you know it works BEFORE you sign a contract.
• Will it be accurate? By running a POC on your actual reconciliation data, you will not only see if it works, but show you the actual accuracy of matching you should expect.
• Can I depend on a fast deployment? Machine learning systems make it simple for you or your staff to set-up new reconciliations without the cost or time of a solutions consultant. Using a prior period already matched reconciliation to train the system, you will be up and running in hours, not months.
• Will it really save me time and provide ROI? A machine learning + cloud based software approach will put you in the driver’s seat by seeing a POC first-hand and its reconciliation capability using your data before you buy. This will give you a clear perspective on how your day will change as you shift your time away from manual reconciliations and towards higher value-add tasks.

Ready to Try a Completely New Approach to Account Reconciliation?

‍Finance and Accounting teams are under intense pressure to close more quickly and reconcile more accounts and more data faster and with greater accuracy. Or maybe you are simply interested in learning more about the next generation finance technology stack. In either case, you owe it to yourself to explore an AI-Machine Learning approach to account reconciliation. Not only will you reduce risk in the financial close process, but you will gain hours of your life back so that you can focus on higher value work or just go home at a decent hour and not work weekends.

Click here to watch a product demo video on Sigma IQ and/or schedule a demo to see how your account reconciliation process can be automated.