Machine learning is on track to transform enterprise as we know it.
A recent survey conducted by MIT Technology Review and Google Cloud shows that machine learning has become an IT priority for the vast majority of organizations, with 60% already using the technology to some extent.
Another 18% say they plan to implement it within 12-24 months. Only 5% of respondents reported that their companies weren’t interested in machine learning.
The survey targeted executives and enterprise developers, with some vice president-level leaders mixed in.
Tech firms are well represented, but so are the financial and business industries.
The size of respondents’ organizations is equally diverse: 48% 50 or fewer employees, 22% 50-1000 employees, and 32% 1000 or more employees.
Aside from the diversity of respondents, the MITTR survey is particularly interesting because of its comparison of current machine learning users to those planning in implement machine learning in the near future.
Placing the groups side by side provides insight into how ML is typically integrated into an organization’s digital strategy over time.
For example, Machine Learning planners tend to push towards applications such as recommendation engines. 56% of executives list recommendations as a primary goal while creating their Machine Learning strategy, along with text-mining (55%) and personalization (47%).
These are the strategies most likely to generate a sufficient ROI to justify further investment.
Companies with mature ML programs express slightly different priorities. They already have a working strategy to build on, so they can widen their gaze. Their objectives for next year are data security (44%), localization and mapping (43%), and targeted marketing (39%).
ML planners relegate those to their five-year plans.
The two groups share some interests. Predictive planning (53%), automated bots (40%), and smart assistants (37%) were equally popular among all the executives surveyed.
Machine Learning in Practice
Before talking about how machine learning is being used, it’s useful to first make a distinction between ML and Artificial Intelligence.
Artificial Intelligence is an umbrella term. It refers to the idea of a computer (or any machine) behaving in ways a human would consider intelligent.
Machine learning is one of those ways: an application of AI in which programs are encouraged to correct and learn from their own mistakes when exposed to data.
The most profitable enterprise uses of ML involve classifying unstructured data and using it to produce business insights.
Here are a few popular applications, listed with their average rate of adoption from the MITTR survey:
Text mining and emotion analysis (both 47%):
These techniques overlap so often it’s easiest to list them together.
Text mining is the use of machine learning to take chunks of text from emails, social media posts, or other sources and analyze them to determine the subject.
Emotion analysis (also called sentiment analysis) estimates the mood of a text’s writer.
These two disciplines are combined very effectively to produce insight about how customers interact with a brand online as well as accurately directing emails to the proper departments without the need for human interference.
Image recognition, classification, and tagging (47%):
Being able to identify the subject of an unlabeled picture is one of ML’s core strengths.
Imagine an algorithm that could sort and tag incoming pictures of damaged merchandise for insurance claims or detect whether pictures uploaded to a social media page violated community guidelines.
Natural Language Processing (45%):
NLP is the backbone of the advanced chatbots used for online customer service, but that’s not its only enterprise application.
A scientist can sort terabytes of online papers using NLP to create a customized list of suggested sources for their research.
Likewise, the time it takes lawyers to find relevant case precedents is drastically reduced.
E-commerce brought in $394.9 billion in the United States last year, growing 15.1% from 2015. At the same time the average attention span of an online shopper dropped 30%.
Smart machine learning strategies by companies like Amazon and Netflix have conditioned consumers to expect to find their product they want without scrolling through multiple pages.
Implementing a machine learning-enable recommendations engine is a good first step to holding a shopper’s attention through checkout.
ML programs can also apply affinity analysis to suggest additional items that go along with products already placed in the cart.
Real World Results
The true measure of success is how machine learning performs in the real world. What kind of ROI are early adopters seeing? Can machine learning really deliver on its promises?
The answer to both questions seems to be “yes”. While machine learning has a long way to go before realizing its full potential, it is advanced enough now to grant a significant competitive edge to its users.
Take a look at some of the most common enterprise goals of machine learning programs:
- Better data analysis and insights (50%)
- Faster analysis and more timely insight (45%)
- Improved internal efficiency (39%)
- Better understanding of customers (35%)
When early adopters of machine learning were asked about the actual results they have experienced so far, here were their answers:
- Better data analysis and insights (45%)
- Faster analysis and more timely insights (35%)
- Improved internal efficiency (30%)
- Better understanding of customers (27%)
The gap between expectation and reality is surprisingly small. It seems more impressive considering that machine learning is showing results faster than many other tech trends.
28% of users with mature ML programs have established ROI, and a third of companies with fledgling ML programs say they’re already showing signs of doing so.
ML users attribute the accelerated profitability in part to increased productivity.
While machine learning does have a learning curve, once the systems are in place they free human employees to focus on higher priority projects.
The lower error rate of ML programs also saves time that would otherwise be spent finding and correcting mistakes.
A recurring theme throughout the survey is the belief that machine learning will give companies a significant competitive edge over their peers. 26% of current ML users feel it already has for their organization.
Machine learning may be a young field, but results like these show it can stand up to the demands of enterprise.
Could your enterprise business benefit from machine learning services? Contact Concepta to explore your options!