What’s Stopping Small Businesses from Adopting Machine Learning?

This article was originally posted on ScoreNYC

Machine learning is, without a doubt, the future. From Amazon’s recommendations engine to Google Photo’s smart albums, the rise of the algorithms promises us a more streamlined, convenient world.

For businesses, increasingly complex and capable algorithms also promise more profits. Yet for now, it seems that big corporations, such as Alphabet, Amazon, or Facebook are reaping the lion’s share of the machine learning windfall. What, if anything, are small businesses to do? More importantly, will the rise of self-learning machines further entrench large companies at the expense of smaller ones?

These questions are not easy to answer, given that the race for more and better AI continues unabated. But as with other technological advances, like cellphones and computers, machine learning will hit a point where it is widely accessible to the general public. Though we are still some ways from this tipping point, we are fast approaching it. More importantly, small business owners do already have some options at their disposal for leveraging this new technology.

What is machine learning?

First things first: machine learning (ML) is a subset of the wider field of artificial intelligence. Specifically, ML is the ability of a computer to analyze data and draw conclusions and to improve its performance along the way. The key to machine learning is that it generally doesn’t require human intervention; instead, it evolves and improves on its own.

Here’s an example: IBM’s Watson, a line of AI products with specialized variants for healthcare, finance, and other areas. Watson’s Oncology bot, for instance, specializes in processing and analyzing terabytes of cancer research, and from this, can generate personalized courses of treatment, and extract relevant insights from medical data.

The key to machine learning is data. A robust algorithm will sift through reams and reams of data (more information than a human can possibly process) in order to come up with solutions. Take Amazon’s recommendation engine, one everyday example of machine learning that nearly everyone has encountered. Responsible for creating close to 35 percent of Amazon’s revenue, the recommendation engine is simply the digital carousel that offers related products that consumers might be interested in.

For such a profitable feature, Amazon’s recommendation engine is surprisingly simple. It’s built on a handful of elements, including past purchases, items left in your shopping cart, products you’ve reviewed or liked, as well as similar goods that others have bought. This information is then collated, analyzed, and presented to the user. It’s important to list relevant, appealing recommendations: unrelated suggestions have a tendency to turn off customers, annoying them rather than encouraging them to purchase more.

What’s stopping small businesses from adopting machine learning?

For small businesses, the key benefits of machine learning are legion. Aside from higher profits (thanks to better data analysis and increased customer engagement), jumping onto the machine learning bandwagon also comes with serious early adopter advantages. After all, in almost every new technology, those who are first to the party tend to improve not just their revenues, but also their market positions.

The issue, however, is reaching that point. Even though a wide range of machine learning services are now available, small-to-medium sized businesses (SMBs) may not yet feel comfortable venturing into the AI arena. After all, SMBs aren’t always on the cutting edge of technology, given their limited resources; in one survey, researchers found that less than 25 percent of small businesses have a website — while only 56 percent have implemented mobile-friendly responsive design, critical for a world dominated by smartphones.

A skills gap may be at least partially responsible for the lag in AI adoption among small businesses. For instance, only 20 percent of executives said their organizations have the technical abilities required for AI. Yet the hardest problem to address isn’t necessarily a lack of capabilities (which can be solved by off-the-shelf commercial products), or even hesitance on the part of leadership (which can be settled with training and exposure). Instead, it’s the thorny question of how to integrate AI into the business. To best understand how to use ML, executives must ask themselves difficult questions, such as: what parts of our business can be automated? What simple (but tedious) work can we take off our employees’ hands and give to machines? What areas of our work are the most prone to error?

Where does machine learning fit in?

For now, machine learning and AI excel at tasks that are rote, routine, and require lots of attention to detail — a combination of requirements that will stymie even the most capable flesh-and-blood employee. One McKinsey studycharacterizes automatable duties as highly repetitive and predictable: think filling data into spreadsheets or trawling reams of data to produce insights.

For companies trying to enter the ML space, it might be best to start off with a specific area of their business, running a pilot program of sorts with an off-the-shelf commercial solution. For example, communication analysis software Monkey Learn can integrate with Google Sheets, Excel, and CSV without any extra coding. From there, Monkey Learn can analyze product reviews, classify and archive inbound emails and support tickets, and from all of this, generate user reports.

For those who are slightly more hands-on (and willing to try riskier approaches), it may be possible to use open-source software, often with the help of outside experts, to craft a custom solution. This step is more practicalfor a medium-sized business with a straightforward goal, a defined budget, and a time frame. Since plenty of ML powerhouses have released their software to the public, either for free or a small fee, this is more practical than ever. Amazon Web Services recently released several frameworks, including Comprehend, a natural language processing service (which can apply advanced pattern recognition to text) and DeepLens, a software that applies neural networks to image analysis.

As an example, consider a business which wants to incorporate chatbots to assist customers with basic support questions, funnel them to the right department for human intervention, and even convert leads to consumers. Obviously, human conversation is a very nuanced, complex form of communication — and one that robots are only now beginning to grasp.

Yet creating an intelligent, virtual program to help triage your customer support and sales departments is now a very real possibility. Just last year, Google released a chatbot analytics platform, allowing companies to track metrics like active users, user retention, and session length, all to get a picture of a chatbot’s effectiveness — and to tinker with it where necessary. More importantly, companies don’t have to build a chatbot from scratch: open-source architectures and off-the-shelf chatbots can apply top-notch machine learning to solve their pain points.

Ultimately, applying machine learning to your business is a tricky, but entirely possible (and profitable) strategy. While it may seem that only large corporations with endless wallets can use ML and AI, the fact of the matter is that a host of startups offer ready-to-use solutions for a wide array of businesses. In fact, the more daring (and well-resourced) can even turn to coding like TensorFlow and, with the help of experts, build their own algorithms.

Within a few years, it’s very likely that ML will become even more mainstream than it already is, spreading beyond tech-heavy applications like marketing or online retail. So why not get a head start?