By John O'Rourke April 17, 2019
Enhancing Finance with Machine Learning: CEO Tom Shea’s Insights
Artificial Intelligence and Machine learning are hot topics in the world of Finance and OneStream Software has been actively researching these technologies and how they can be applied in corporate performance management (CPM). OneStream CEO Tom Shea has been a keen advocate of these technologies and has spoken about them at our Splash user conference and other events.
Based on the popularity of these sessions and positive feedback afterwards, I sat down with Tom to discuss and further understand the importance of Machine Learning and how it can transform business today and, in the future, specifically around CPM processes such as financial close, reporting and planning. Here's what he had to say.
John: Machine Learning is not a new concept and yet it seems to be one of the hottest topics in technology today. What exactly is Machine Learning and why is it so popular now?
Tom: Machine Learning is a branch of Artificial Intelligence, or AI, that is specifically focused on software that has decision-making capabilities based on recent experiences and past trends. It goes beyond traditional rules-based programming, leveraging statistical algorithms to learn and get smarter over time, retraining itself as it gains more 'experience'.
As you note, it's not a new concept but it's finally hitting the mainstream with commercially viable
applications that autonomously improve business. The combination of today's raw computing power with massive amounts of information, sometimes called Big Data, is fueling machine learning along with recent advances in complex neural networks. These deep-learning models allow applications to analyze enormous quantities of data points, creating new efficiencies beyond what linear models can provide.
John: As Machine Learning becomes more accessible, how will this benefit businesses? Where does Machine Learning make the most sense?
Tom: Actually, Machine Learning has already greatly benefited the financial services industry with applications such as fraud detection and anti-money laundering. Analyzing colossal volumes of data for patterns that indicate criminal activity has been an early success for Machine Learning – the pilot project, if you will, that is driving interest in other areas.
But don't get drawn into exaggerated scenarios where Machine Learning automates everything and replaces the workforce. Expect instead to see Machine Learning applications for tasks that are tedious and time consuming, allowing people to focus on more strategic projects. Monitoring medical device data in healthcare and alerting patients and medical staff when something requires attention, or even assessing credit and insurance risk by reviewing applicants and historic data are use cases that provide incredible value without displacing any human beings.
Machine Learning is also good at understanding unique customer preferences and improving overall service. In retail and entertainment, we are seeing an improvement in recommendations and a move toward more personalization – from product advice on Amazon to suggestions for what to watch next on Netflix.
John: How about Machine Learning in corporate business processes? What do you see as opportunities to improve efficiency?
Tom: Every department can be impacted by Machine Learning. Sales and Marketing teams can use it to predict customer churn and proactively target promotions. HR departments can leverage Machine Learning to enhance recruiting and retention of talent. Operations will get smarter at resource deployment, scheduling and purchasing.
John: What about Finance and CPM processes? How will Machine Learning improve the financial close and reporting process, or planning and budgeting?
Tom: CPM is ripe for Machine Learning to help deliver new efficiencies. At a technology level we are moving toward autonomous CPM with our X-Scale initiative. "Autonomous CPM" may sound ambitious but consider autonomous driving. Self-driving cars are no longer just a futuristic concept, in fact, autonomous driving is poised to radically shake up the automotive industry as we know it. Data sensors generating millions and millions of data points, real-time analytics and Machine Learning are critical to enabling the capabilities that allow a car to drive itself.
Now think about CPM. If we gather enough data points to learn about system performance under certain conditions, we can react and optimize computing resources – including the infinite scale of the cloud - scale computing resources up and down when needed to optimize performance. WithSmartCPMTM there are also opportunities to speed account reconciliations, identify errors in data integration, automate account mapping, and improve the accuracy of financial and operational planning. It's all about leveraging diverse data sets – financial data, operational data, external economic or industry data – and getting them into the right structure to applying Machine Learning algorithms, then analyzing and acting on the results.
When you think of the CFO's team, most would like to have more time for the strategic tasks but get bogged down in all the tedious but necessary steps for a timely and accurate monthly, quarterly or annual close. Leveraging technologies such as Robotic Process Automation, Artificial Intelligence and Machine Learning can help them shift their focus to value-added analysis and on making informed decisions that can help a business get ahead.
John: Where is OneStream Software on incorporating Machine Learning (ML) into our platform and solutions?
Tom: OneStream is taking a very strategic approach to how we are leveraging AI and ML in our platform. Our goal is to make Machine Learning accessible and applicable in ways that make our customers and users more productive, and to let data science teams work with the tools and engines that they prefer. OneStream's strategy is to integrate predictive analytics & machine learning into two elements of our unified platform and user experience.
First - autonomously scaling our platform. With the OneStream XF 5.0 release, we provided the foundation for "Intelligent Scalability" with the first phase of our X-Scale architecture. The current release includes environmental sensors and "intelligent BOTs" which gather and review data regarding server usage, task queue, database, application servers, logons etc. BOT servers then use rule-based programming to perform smart load balancing, enhancing platform performance and scalability in both on-premise and cloud deployments.
In subsequent releases the X-Scale architecture will include rule-based and schedule-based scaling of server resources in the OneStream XF Cloud, then Machine Learning algorithms will be trained to scale up or down server resources based on usage patterns (e.g. during month end close, budgeting season). This will optimize both performance and cost for all customers by allowing them automatically take advantage of the infinite scale of the cloud as needed and to scale it back down when not needed.
John: Wow, that's exciting and unique – I haven't heard of other CPM vendors providing that type of scaling capability in their cloud platforms. What's the second area of focus for ML?
Tom: Second is in planning and forecasting. OneStream's vision is to enable users to leverage predictive and/or machine learning models as a non-biased forecast scenario in planning, budgeting, and forecasting.
Customers can already leverage OneStream's Sales Planning and Thing Planning applications from the XF MarketPlace to facilitate predictive forecasting using regression and goal seek functions. But OneStream has also developed a complete Machine Learning framework solution that we are planning to release later this year, along with data scientists we are hiring to deliver and enhance this solution properly.
This solution, which we call ML 123, will help solve one of the largest challenges data scientists have – which is to translate their work into something the business can consume and utilize in driving business performance.
ML123 will provide a framework for data preparation and integration with leading data science engines (initially Azure ML, Amazon ML, IBM Watson, etc.) and analytics for business and financial planning. For customers who have made investments in data scientists and technology, ML123 will make it easier to bring together all the data that data science teams need to 'work their magic' and to bring the results of their work into the OneStream platform, where we make it easy to consume and utilize the output for advanced analysis, forecasting and to make critical decisions.
John: This sounds powerful, and different from other vendors who are attempting to build a "one size fits all" approach with building predictive and ML-based forecast algorithms into their platforms.
Tom: That is exactly right John. With Machine Learning, there is really no such thing as a "one size fits all" approach. That is because every business and every part of a business is unique. For example, the ML model that serves one product or service best is likely not going to work for another product, or even the same product in a different part of the world. The weather is different, consumer preferences can be different, there could be different macroeconomic impacts playing out and different seasonal patterns. Our goal with ML 123 is to enable companies to create multiple, optimized ML models – apply them to key business problems and extend these insights across the enterprise - leveraging their investments in ML and the power of our platform.
John: Thanks Tom. I'm sure our customers will appreciate your insights and the time and thought we are putting into the application of Machine Learning at OneStream. We look forward to hearing more about this in the near future.
To learn more about OneStream's plans to make SmartCPM even smarter join us at our Splash user conference in May. Learn more and register here.