Doing more with less is the mission of every IT executive. No matter what resources or budgets are allotted, we continuously search for new ways to increase efficiency, speed, and agility.
Digitalizing and automating processes is a must-win for most enterprises. But traditional automation delivers the biggest benefits when companies can precisely define rules and processes. When data or processes are even a little fuzzy, precision suffers – and automation is less effective.
That’s why many business leaders are looking to machine learning to support task automation. Wikipedia defines machine learning as “a field of computer science that gives computers the ability to learn without being explicitly programmed.” It’s a type of artificial intelligence or digital labor that “learns” and improves based on experience.
With machine learning, organizations can take manual processes, apply algorithms, and make sense of vast amounts of historical data that couldn’t previously be used effectively in automation processes. It is especially useful where business rules are complex or there are high volumes of unstructured data.
Technology executives see the value. In the 2017 Harvey Nash/KPMG survey of CIOs:
- 62% of respondents from larger enterprises are already investing or are planning to invest in digital labor (which includes artificial intelligence, automation, and robotics)
- 27% of those surveyed expect machine learning to improve quality
- 24% of executives say machine learning is most effective for increasing efficiency
Automation’s impact on shared services
Many companies use shared services to address those fuzzy processes that cannot be automated effectively. For example, finance organizations often use shared services organizations to handle invoice processing because it is a non-exact process that can be difficult to automate.
Typical invoice processing problems include mismatches between remittance advice and customer data, currency mismatches, partial payments, insufficient remittance advice, and single payments for multiple invoices. Traditional rule-based automation does not yield significant improvements in invoice processing times, productivity, or efficiency.
SAP uses a global finance shared services organization to match incoming bank statements (payments) with open invoices (receivables). This team, located in Singapore, processes 85,000 invoices each year for the entire Asia-Pacific region. You can imagine the number of hours required to address invoice matching issues.
To streamline these tasks, we deployed a new machine learning application that automates the invoice matching process. The software identifies more than 40 features – such as amount difference, data difference, and misspelled words – that can be used to match incoming bank statements with our receivables, which correspond with invoices. We then used historical datasets from our database to develop a prediction model.
We used the technology to test the processing of approximately 7,500 invoices for customers in Singapore. In this project, we were able to simulate the closure of over 50% of our invoices against incoming bank statements. The remaining invoices can be closed manually using proposals generated by the machine learning model – further simplifying the existing manual process.
The machine learning technology offers quantifiable value to our organization, including:
- Approximately 20 hours per month savings on manual reconciliation efforts by our finance organization in Singapore
- Ability to clear outstanding invoices almost immediately, reducing the daily sales outstanding rate
- Improved accuracy and consistency of the reconciliation process
Better business value through machine learning
Perhaps most importantly, we see opportunity to gain new value in the years ahead. Based on our experience, we can envision further automating as much as 50% of our shared services tasks in the next five years.
Machine learning can help companies become more agile, competitive, and efficient. In shared services organizations, the technology offers new opportunities to process simple issues quickly and accurately, further streamlining operations. Freed from these time-consuming manual tasks, financial experts can concentrate on more complex issues. And financial organizations can use hard-won expertise to become centers of excellence that deliver enhanced value to the business.
I plan to continue sharing stories about SAP’s digital transformation journey, and I encourage our readers to do the same here in the Digitalist Magazine. If you have a story to tell about your organization’s digital transformation, please contact Jean Loh, jean.loh@SAP.com, who manages the CIO Knowledge section of the Digitalist. We look forward to hearing from you.