Our Finance and Global IT teams collaborated with Internal Controls to make the pre-close variance process and reporting digital, automated and intelligent. The vision was to create a single standardized and centralized pre-close variance report that was easily accessible across controllership and provide the ability for controllers to add commentary and document their review. This was the starting point with the intent to add more capabilities over time.
Given Accenture’s global size and scope, the project was a succession of progressive, agile development phases over two years:
Creation of an online report
The first phase was the creation of a digitized, global pre-close variance report with the ability to add variance commentary manually. This report eliminated the need for data downloads from the SAP system into Microsoft Excel, manipulation of data, and multiple iterations of the local download. In key month-end close days, the data refreshed regularly.
This report became the same for global and local controllers, providing relevant views for each user group. Features included internal control validation, dynamic threshold analysis, and the ability to add comments at different levels, such as company code and the general ledger account levels. The commentary framework was designed to allow for future automated analytics.
Addition of automated commentary
This phase of development added automated commentary to the pre-close variance report. This capability populates commentary content as a starting point for controllers to validate, and if necessary, revise. The aim is not to replace all manual commentary, but to ultimately produce up to 95% of the content, and to guide users into more informed analysis. The benefit of auto-populating commentary is that it provides an unbiased, high-quality baseline of data. An early alert system was also developed to enable users to act on unexpected items in a timely manner.
Automated commentary models are designed to work for and with the users, providing support by 1) automating the explanation of common and expected behavior in variances; 2) providing data-driven insight regarding general and anomalous patterns; and 3) guiding controllers when exploring large variances and unexpected entries.
For example, in the case of an automated comment for travel air fare, the artificial intelligence engine looks at a large number of drivers that would make up a variance in a particular month’s air fare, such as headcount increase, number of working days, holiday periods, airport departure/destination locations, and a comparison with the previous year’s monthly numbers, among other factors.
Addition of variance insights
Our project team then enhanced the pre-close variance report with added capability that provides end users with more insights on variances by enabling them to drill down into the lower-level line item source for a specific variance. This capability eliminates the need for users to leave the report and go into the SAP system to find the cause of a variance or to validate the automated commentary. All needed information is now in one report.
Finalization of auto-learning of existing patterns
The natural language generation (NLG) engine was continuously enhanced using two years’ worth of historical commentary and manual inputs from controllership review. The development team is planning to add new patterns and to enable the engine to continuously auto-learn patterns.
The pre-close variance report was rolled out in waves that started with a soft go-live in a small number of Accenture countries to pilot and refine. It is being fully deployed Accenture-wide over the course of 2022.
