Machine learning has emerged as one of the biggest and most promising technology trends today, poised to greatly transform our lives and how we do business. We are already seeing early applications that provide smarter customer recommendations for online services such as Netflix or Amazon, the rise of virtual assistants like Siri, Alexa and Cortana, and improvements in healthcare such as better image scanning to detect cancer. Machine learning is the driving force behind autonomous vehicles, too.
A branch of artificial intelligence (AI), machine learning is essentially software that can make decisions based on experience and without the need for traditional rules-based programming. It uses statistical algorithms to learn and get smarter over time, retraining itself the more it "experiences."
The convergence of three other technology trends is pushing machine learning from concept to commercially viable applications that are autonomously improving business:
- We finally have the raw computing power to enable the volume of transactions or "experiences" required for machine learning.
- Massive amounts of information – Big Data, if you will – including shared external data from Internet of Things (IoT) applications gives machine learning the broad context it needs.
- Advances in complex neural networks, or deep learning models that can analyze enormous quantities of data points are creating efficiencies beyond what linear models can provide.
The ability to go beyond detecting patterns to proactively adapting and optimizing a solution is what makes machine learning valuable. Financial institutions have used predictive analytics for more than 20 years to combat fraud. Where predictive analytics software helps detect troubling new trends that could be fraudulent, a machine learning application can act to prevent potential fraud. Plus, it continuously improves, getting ever smarter at prevention.
Machine learning is especially effective for applications that process vast amounts of data, particularly where traditional linear models can be limiting. With the ability to process unstructured data, machine learning can discover patterns and correlations that were previously undetectable. Beyond fraud prevention and other obvious uses, machine learning is also automating tasks that can be tedious, allowing IT staff to focus on more strategic projects.
Early successes have been largely in the consumer space, but we are now seeing business applications for machine learning, with great potential across multiple business functions:
- For the sales team, machine learning will deliver more insight from Customer Relationship Management (CRM) systems and a deeper understanding of customer churn and buying trends to ultimately improve customer service and shorten sales cycles. This will be competitive advantage to those who adopt early and get it right.
- Machine learning can help bridge the gap between sales and marketing by determining the correlation between marketing programs and unit sales. Further, it can help with more effective customer segmentation to support marketing campaigns, promotions and efforts to close on sales.
- Expect HR departments to leverage machine learning for recruiting and retention of top talent.
- Operations will get smarter at planning, resource deployment, scheduling and purchasing as a result of machine learning.
- The finance department will deploy machine learning to help manage cash flow, speed account reconciliations and improve overall financial planning.
But while machine learning offers tremendous benefits, there are many challenges in implementing it as well. Any machine learning application is only as good as the data to which it has access, as well as its overall modeling structure to access, learn and act. Usability is another key factor to enable broad use across the business.
The real race to provide the best machine learning applications perhaps lies in who gets the best data scientists. We are already taking great strides in machine learning, but this is still very much the early stage in what is largely uncharted territory. Data scientists today are developing best practices that could define the future of machine learning and other forms of AI. Beyond building business functionality and optimizing algorithms, how we handle outlier results, overfitting or false positives will determine success.
Similar to the early days of the Cloud, the Internet or even computing itself, it is an exciting time for machine learning. Given the many exciting applications machine learning promises, expect to see more solutions coming to market soon.
