How to leverage Machine Learning applications in Finance

Artificial Intelligence

,

Financial Management

How to leverage Machine Learning applications in Finance

Isabel Calvo | Apr 18, 2019

In recent years we have witnessed the emergence of new technologies in the finance industry, such as Big Data and Artificial Intelligence, which in turn has led to the rise of “Fintech” companies. The latter employ recent technological advances to improve their processes, products and financial activities

The way of approaching business and financial operations in the 21st century has completely changed the business panorama, making it necessary for companies to quickly join the innovation race.

The automation of standard processes has always been the top priority for finance departments. This has evolved to a higher level. Thanks to technologies such as intelligent software and Machine Learning applications in Finance, it is possible to effortlessly achieve the automation of demanding processes such as bank reconciliation.

How to apply Machine Learning in Finance

What is bank reconciliation?

Bank reconciliation is a process that enables the company to verify if the activities that the bank records in its account coincide with the operations that have been effectively performed and entered in accounting. In essence, it is about identifying a bank charge or payment (using the statement that the bank sent) with its accounting source: the item listed in the financial entry that was paid or charged for.

Bank reconciliation is a fundamental internal tool for any company that wants to know its cash flow and liquidity situation, or if it wishes to have a control mechanism against fraud or the mismanagement of the company’s cash outputs and inputs.

In reality, this is not a simple task. Errors, omissions and duplicate entries can exist that unbalance the reconciliation and require a lot of time to clarify. This is because reconciling banks often resembles a work of art in a sense, where we need to note by hand – literally by hand – and with ruler, paper and pencil, hundreds of items of unknown origin and match them to transactions on the bank statements.

SAP Cash Application and Bank Reconciliation

To make it easier for us and to not have to worry any more about creating reconciliation rules in our systems, or about having to waste countless hours manually mapping entries, SAP is incorporating SAP Leonardo's SAP Cash Application into its product suite:

“First there was manual reconciliation,
then came automatic reconciliation,
and now SAP is launching the SAP Cash Application”

What is the SAP Cash Application?

SAP Cash Application is the integrated SAP solution for SAP S/4HANA. It is focused on improving the bank reconciliation process and cash flow management through the use of artificial intelligence methods and Machine Learning (technology contained within SAP Leonardo).

Machine Learning is a branch of the artificial intelligence that was created as an effective technique for automating processes. It enables the identification of hidden patterns between the information and knowledge contained within the process, in this way learning the extracted data without being explicitly programmed for it.

The term Artificial Intelligence isn’t entirely new. In fact, it was coined by John McCarthy around 1956. However, we're now at the moment where computers have become powerful enough to analyze immense volumes of data (Big Data), thus enabling data scientists to develop and use models based on this information.

With the SAP Cash Application, SAP pledges to achieve new levels of automation and productivity in the clearing of payments received by using artificial intelligence and Machine Learning in bank reconciliation.

SAP Cash Application and Bank Reconciliation

The majority of companies have reconciliation software that enter payment transactions automatically and that guarantee the mapping of the latter using previously created rules. This software makes it so that the finance team only has to manually intervene in the process when there’s an exception, such as in the case of payment data omissions, written errors by the company or bank, duplicate entries; or in cases where several invoices are paid in one transaction.

This is where Machine Learning comes into play.

A software enabled for machine learning “learns” in a way that is similar to how a human would: through experience, observation and historic data. It utilizes self-learning algorithms to detect patterns, recognize contexts and make predictions, with the objective of mapping bank payments from the electronic statements with the accounting items from the accounts receivable department.

How does the SAP Cash Application process really work with Machine Learning?

A brief phase of “familiarization” is an essential preliminary step before one begins the process of bank reconciliation itself. Here, the specific model for each client is created and trained (each company has its own rules and commercial scenarios) and will be accepted as a basis for mapping and reconciliating items in the future.

This model – as we have commented previously – draws from the client’s historical information: bank statements, accounting documents, payment notices, remittances, client maestros, bank maestros, one’s own bank information, payment information, bank services such as Lock Box and information and examples of past offsets (at a minimum the model needs to be trained with 5K of information).

Once all the data is collected, it is sent to the SAP Cash Application services and gathered by the Machine Learning training engine. The latter, based on algorithms, internally chooses and sorts the criteria and selection values, resulting in client mapping and the optimal categorization model.

How does the SAP Cash Application process really work with Machine Learning?

The actual implementation of bank reconciliation in the client’s system begins once the basic model has been trained.

Therefore, the process in essence really begins when the client pays their pending invoices, in such a way that the accounts receivable department should receive notification of this payment.

Often, when the clients pay their invoices, they send a payment notice indicating which invoices they have paid. SAP Cash Application has the ability to extract the invoice information that needs to be cleared from these remittances, in accordance with the delivery configuration instructed by the client.

Once the information has been collected from the bank statements and accounting invoices, the data is transferred to the engine for the SAP Cash Application, which is based in the SAP Cloud Platform. This gives back the proposed matches and their confidence level for success. This confidence level is employed by the user to evaluate if they prefer to automatically reconciliate the engine results, thus automatically eliminating the items and proceeding to enter them; or if, on the other hand, they prefer that they only be suggestions pending review.

All of this process is cyclical, replenishing the engine with each new delivery, learning the specific behavior of the client/country from past actions and creating new data patterns for resolving all possible exceptions.

Key Functionalities of the SAP Cash Application:

  • Proposes open debtor items as a possible mapping method with regard to the bank statements.
  • Enters and automatically clears the items.
  • Using condition filters, it enables its implementation through programmed Jobs.

Steps for the Bank Reconciliation Process in SAP:

  1. The bank statement is uploaded electronically into the system (transaction FF_5) or manually (transaction FF67).
  2. The standard reconciliation rules are implemented for SAP S/4HANA.
  3. The execution of the Job for the SAP Cash Application is programmed. This sends the data from the open accounting items to the SAP Cash Application services (where the trained reconciliation model is hosted) with the new electronic bank statements.
  4. With Machine Learning, the SAP Cash Application sends back mapping proposals to the SAP S/4HANA system.
  5. Reconciliation is executed automatically, entering and clearing open items, provided that the mapping proposals have a degree of confidence that's equal or superior to the system's predefined tolerance.
  6. The bank statement is reprocessed and outstanding payables are manually reconciliated (FEBA and FEBAN transactions).

Benefits of the SAP Cash Application

Automation and control
  • Automatically learns historic data: The model automatically trains itself with historic data and creates its own consolidation rules.
  • Learns from user actions: The model adapts and relearns from the user’s interactions.
  • Does not require maintenance: Depending on how the user interacts with the process, the model gives feedback and modifies the rules according to the actions taken by the user.
  • Reduces errors and improves process control: Leaves manual errors by both the bank and finance department as pending reconciliation, thus enabling their rapid identification and correction.
  • Greater precision in reconciliation.
  • Flexibility in the solution: Allows for the adaptation of the solution to the changing needs of the client, opting if they prefer for the trained “compound” model that provides suggestions for reconciliation, but does not implement the mapping without prior supervision and action by the user.
Time and cost reduction
  • The need to manually create reconciliation rules is eliminated: The model is trained to create the client’s own mapping rules. It achieves high automation rates without having to manually adjust the system.
  • Reduces the complexity in resolving imbalances: Leaves manual errors and new items exclusively for mapping that won't reappear during user actions.
  • Enables the finance team to concentrate on more strategic tasks: A much faster reconciliation saves time as well as the finance team’s attention so that it can perform more complex and strategic tasks.
  • Increase in the finance department’s efficiency.
Improves KPIs
  • Improves average collection period (ACP): Improves the resolution and analysis of disputes for unpaid charges by being able to really know which invoices haven't actually been paid for.
  • Improves liquidity and working capital: Enables more efficient control over the company's actual liquidity position by making the mapping of items faster for collections and payments made.
  • Enables you to provide a faster and more efficient service to the client.
Integrated with SAP S/4HANA
  • Cloud or On-Premise: It adapts both to SAP applications in the cloud and to traditional apps (on-premise).
  • 100% integrated: Works instantaneously and automatically with the implementation of SAP S/4HANA.
  • Supplementary to the standard rules: It can be used in combination with the standard rules that are already defined in client systems.
  • Keeps the current flows and workflows defined in the system.


Interested in learning more?

Start implementing Artificial Intelligence and Machine Learning to improve your Finance processes! View our services and solutions for Artificial Intelligence.

LEARN MORE

Subscribe!