Special Feature
Part of a ZDNet Special Feature: How to Implement AI and Machine Learning

Five ways your company can get started implementing AI and ML

Are you looking for ways to harness the power of machine learning and AI for your business? Here are five tips for beginners who want to gain meaningful insights and predictions from their data.

ai-ml-five-ways.jpg
Image: Getty Images/iStockphoto

FREE DOWNLOAD

Special Report: How to Implement AI and Machine Learning

You can download all of the articles in this series in one PDF. It's free to registered ZDNet and TechRepublic members.

Many businesses today understand that AI and machine learning -- which uses data to make predictions -- is the way of the future. It is the fuel behind image recognition, speech processing, translation, and other tasks that have business implications for marketing, customer service, and many other disciplines. For instance, according to a 2015 report by McKinsey, "predictive maintenance" by manufacturers could save between $240 billion and $630 billion by 2025.

Although the significance is clear, dipping your toes into AI can be a daunting task. So how can businesses get started? Here are five ways, according to CEOs and AI experts who have gone through the process.

SEE: Baidu launches open AI platform to give businesses an easier on-ramp than Google or Microsoft

    1. Learn how machine learning can help your company

    "When preparing to use machine learning, the first thing organizations must do is train lead engineers to have a solid understanding of the technology; how it works and what advantages it can deliver," said Chris Rijnders, CEO and co-founder, Cogisen. For example, Boeing has set up a joint lab project with Carnegie Mellon, he said, "so that its engineers can understand its potential impact in every aspect of design, manufacturing and maintenance." This demonstrates how critical education should be when applying machine learning to complex environments, he said.

    2. Research other businesses already using AI and machine learning to determine parallels

    "AI and machine learning are not yet in the DIY category," said Fabio Cardenas, CEO of Sundown AI. "It's all still very technical." So, it's worth finding out what other businesses have similar goals, and how they have addressed the issue.

    3. Choose a platform

    With Amazon, Baidu, Google, IBM, Microsoft and others all offering machine learning platforms for the enterprise, there is no obvious place to start. Many of these options are similarly priced, and aimed at beginners. Check out the individual articles on these platforms in this special feature to help you decide if one of them is right for your business.

    4. Create a strategy

    Data science companies like Boxever can help businesses deploy AI -- for example, by addressing a question like, 'How can AI improve marketing?' AI could help you make predictions about what happens when customers open an email, for example, based on previous experience. This is an easy way to integrate AI into current operations, said Dave O'Flanagan, Boxever's CEO and co-founder, because it helps "build trust."

          "We had to introduce a lot of controls on rules to be able to allow organizations to treat the output or deploy their own strategies themselves," O'Flanagan said, "and then put their strategies alongside black box or AI strategies to be able to get comfortable with the concept of a machine making decisions about what kind of information to present to a customer."

          5. Create an implementation plan

          Before you can get started deploying your product, you need to think about a plan. According to Sundown AI's Cardenas, a multi-region deployment plan using Amazon Web Services (AWS) has a detailed description for users. "Setting up the AWS infrastructure would take a few days, assuming that the web application has been tested on such infrastructure previously," he said. If it hasn't, you'd need to set up "the web app, database and other related infrastructure on AWS, connecting all the components," Cardenas said, which could take a week or two. Additionally, it would require constantly refining the coding for bugs, which would call for additional deployments. Cardenas estimated that the process for a "deploy pipeline" could take another ten days or so.

          Also see

          Newsletters

          You have been successfully signed up. To sign up for more newsletters or to manage your account, visit the Newsletter Subscription Center.
          See All
          See All