You want to democratize AI.
You want to engage your people in the AI process. You want to tap your team’s hidden domain expertise.
Above all, you want to eliminate bottlenecks and get stuff done. Machine learning experts are hard to find and hard to keep. You can’t transform your business with a handful of experts.
So you figure: it’s time to invest in a drag-and-drop tool for AI.
You’ve seen the marketing. Or maybe you saw a demo at a trade show. Drag-and-drop tools for AI look easy. It’s simple. You drag icons to the canvas, connect them and voila! Do it yourself machine learning!
In the back of your mind, though, you may think it’s not as easy as it looks. You should listen to that voice.
Drag-and-drop is not a new idea. SPSS introduced a drag-and-drop tool for advanced analytics way back in 1992. Since then, we’ve seen multiple generations of drag-and-drop tools. Fat clients, thin clients, and browser-based tools running on desktops, servers, and in the cloud.
We’ve had drag-and-drop tools for twenty-four years. So why are we still struggling to democratize? Here’s why:
You still need to know what you are doing. Users must know what to drag and where to drop it. Mature drag-and-drop tools have hundreds of operators. Vendors brag about all the operators they have. Operators are nice, except that users need to know which operator to use first, and what comes next. Have you noticed that vendors who pitch drag-and-drop tools also offer lots of training? If the software is so easy to use, why do users need so much training? Why do your people need to get certified? Why does it take weeks to get users up to speed?
Complex projects have thousands of tasks. Vendor demos show you projects with a few operators. Drag-and-drop tools look great if you only need to drag and drop five things. Have you ever seen the workflow for a complex project? Typical AI projects have thousands of tasks. Try doing that with a drag-and-drop tool. Instead of democratizing AI, you’ll get complaints about carpal tunnel syndrome.
No idiot-proofing. When you democratize AI, you need to build in guardrails. Your team may include the best and brightest people in the business, but everyone makes mistakes. When you empower people with drag-and-drop machine learning, you give people the power to drag and drop something dumb. Or forget something. Are you going to take that risk with mission-critical AI? No, of course not.
You’re still going to need experts. For all the previous reasons, you’re still going to need a team of experts. Business users won’t want to spend weeks building a complex workflow. Moreover, you’re not going to trust novice users to build mission-critical AI with drag-and-drop tools. Instead of democratizing AI, you’re just going to add another application and create a new silo.
A top strategy consulting firm invested heavily in drag-and-drop machine learning tools. They figured it would be smart to put these tools in the hands of their front-line consultants. However, the consultants didn’t have time to master the new software. The firm had to create a new team of experts to build complex workflows.
Incidentally, data scientists hate the tool. They already know how to write code, so they don’t need a drag-and-drop UI. It’s inflexible and scales poorly. Also, those drag-and-drop workflows are hard to audit.
With DataRobot, you drag and drop precisely once -- with your data. DataRobot does everything else. It builds hundreds of complex pipelines, like this one:
By the way, that’s a simplified view. Each one of those boxes may require dozens of operators with a drag-and-drop tool. And that assumes you know what tools you need, and in which order to use them.
Since DataRobot does that work for you, it does it right. You don’t have to worry that a user will forget a step, or do something silly. DataRobot’s built-in best practices deliver expert-level AI every time.
Would you rather prepare a banquet from scratch, or dine out at an excellent restaurant? Do you buy lumber and build your office building or do you just rent space? Would you prefer to buy a new car or assemble your own from parts?
Because that’s what vendors of drag-and-drop software want you to do. Build your own machine learning from a bag of parts.
The choice is simple. You can invest in a modern tool for AI like DataRobot, and jumpstart your AI transformation.
Or you can invest in a drag-and-drop tool. It’s so 1992.
About the Author
Thomas W. Dinsmore serves as Senior Director at DataRobot. Prior to this, Thomas was the Director of Product Marketing and Data Science at Cloudera and ran his own consulting company. Thomas has his MBA in Accounting and Decision Sciences from University of Pennsylvania - The Wharton School.