Getting Started with Machine Learning in the Enterprise
Machine learning (ML) is a subset of artificial intelligence (AI) in which data scientists use algorithms and statistical models to predict outcomes and/or perform specific tasks. ML models can automatically “learn from” data sets to improve their performance.
ML is uniquely applicable to enterprise business use cases across a wide number of industries — for example, credit scoring and fraud detection in financial organizations, and tumor detection and DNA sequencing in healthcare. Given the wide applicability, it’s no surprise that many in the enterprise have already embraced machine learning. According to Deloitte Insight’s 2018 survey of US-based early adopters, 63% were already using ML in their enterprise organizations.
Potential + Timing = Opportunity
As with any breakthrough in technology, machine learning’s incredible potential doesn’t mean much if the enterprise can’t take advantage of it. In a recent research note, however, Gartner® points to four factors currently driving further adoption of ML and other AI initiatives:
- Easier access to high-performance compute resources
- Increasing availability of broad datasets
- Maturing open-source tools and platforms (like Python and Anaconda Enterprise)
- New marketplaces offering pre-built algorithms and other software components
What’s Standing in the Way?
Despite ML’s great potential, the enterprise faces significant obstacles on the path to adoption. These challenges include:
- A rather fuzzy general understanding of machine learning, exacerbated by the fact that people often use terms like ML, data science, and AI interchangeably.
- Confusion about how data science teams (who would drive any ML initiatives) fit within the traditional silos of IT. (Hint: They don’t.)
- Little consensus on how to build an effective internal data science team, which roles are needed, and how to address the skills gap within the organization.
- Resistance to change among internal business analytics and line of business leaders, who may need proof of ML’s unique benefits in order to embrace a new way of thinking and working.
- A lack of best practices for operationalizing data science and ML at scale.
Considering that data science, AI and ML are still-emerging disciplines that represent a seismic shift in how business works, these issues are more than understandable. The good news is, they’re not insurmountable.
Tips for Getting Started with ML in Your Enterprise
- Follow the leaders
Rather than breaking new ground in your first ML initiative, focus your early efforts on proven use cases such as marketing, financial planning, sales or risk management. Other organizations have already proven machine learning’s applicability to these areas of business — capitalize on their experience.
- Tackle a real business problem
Work with the leaders in your lines of business to uncover their intractable pain points, then look for ways in which ML can help alleviate those issues. You’re looking for a relatively quick win as a proof of concept. With demonstrable proof of improving or eliminating a genuine problem, you’ll gain champions across the business and unlock more enthusiasm and budget for your next (and likely larger) effort.
- Don’t expect linear progress
Data science is inherently exploratory and iterative, so your machine learning efforts are unlikely to follow a traditional path from Point A to Point C, D or E. Work with your data science team, whether internal or external, to set touch points along the way so you can understand their insights and progress.
To learn more about machine learning and how to get started, download your free Gartner research note, “Machine Learning: FAQs from Clients.”