Twelve billion dollars. That’s how much investors poured into Artificial Intelligence (AI) startups last year.
The inevitable result: an explosion of AI startups. Crunchbase, a public database of startups, lists more than 7,000 AI vendors -- two-thirds of them less than three years old. Wired magazine says that AI experts can raise capital without an idea, product, or business.
Venture capitalists see the potential of AI -- and so do business leaders. In a recent survey of 1,600 executives, 61% cite AI and machine learning as their top data initiative. Consultant IDC predicts that worldwide spending on AI will reach $58 billion by 2021.
The explosion of new vendors is just one of the things that make it tough to choose an AI vendor. Here are some others:
Fins on a Buick. Legacy software vendors add AI functions to an aging code base and call it "AI." That's like adding fins to a 1950s Buick and calling it a rocket. Legacy software cannot support AI at enterprise scale.
Wizard of Oz. The Guardian documents the rise of "Wizard of Oz" AI vendors. These vendors claim to offer "AI in a bottle." In reality, they employ data scientists behind the scenes who do the work. This practice is surprisingly widespread.
Techno-Babble. AI is a complex technical field, with a language of its own. Executives struggle to communicate with AI experts. Executives want to talk about business. Experts speak about hyperparameterization, deep adversarial networks, and TensorFlow epochs. Finding a common ground is a challenge.
Battle of the Experts. AI is a collection of disciplines: machine learning, image recognition, natural language processing, and so forth. Each of these disciplines has its own language. Deep learning experts speak one language; text mining specialists another. Practitioners who use one open source library may struggle to communicate with those who use a different platform.
Pervasive hype and hot air lead some executives to take a “wait and see” approach. This is understandable. But the smartest companies are already placing their bets on AI; those who wait will be left behind.
The good news: you can distinguish the best AI partners from the charlatans.
Here’s what you should seek in an AI vendor:
Pragmatic Approach. AI has theorists and "opinion leaders" who publish their ideas in business journals. That's a good thing: theory drives innovation. However, many ideas look great on paper but don't work in practice. Look for vendors led by executives with hands-on experience making AI work in large organizations. Your best partners build testing and validation into everything they do, and they build best practices into their solution.
Viability. Will your AI vendor still be around in a year? Or three years? Most startups don’t simply go out of business. Someone else acquires them, or acquires their assets. When someone else acquires your AI vendor, you pay. Look for well-funded vendors with steady headcount growth.
Modern software engineering. Look for an AI platform that is scalable and elastic; runs everywhere; and leverages an open source software foundation. If a vendor’s product is more than ten years old, check under the covers. Make sure that you’re not buying old wine in new bottles.
Support for a broad range of users. We hear a lot about AI "heroes." Some vendors offer products that are FEO: For Experts Only. Others think that AI should be so easy that a caveman can do it. In reality, your organization has diverse users with different needs. Some users want to keep using the tools they have today. Seek a vendor that offers the open APIs and business partnerships that make that possible.
AI workflow support. Executives complain that it takes too long to get AI into production. The root cause of the problem: a broken AI workflow. Rethink the workflow to reduce cycle time. Look for an AI vendor that understands and supports AI lifecycle, and will help you organize for rapid value.
Commitment to customer success. Transforming your business with AI is hard. Some vendors will try to tell you it's easy if you buy their product. Don't believe them. Seek out vendors that understand the challenge and design their offerings accordingly.
Learn more about how to choose an AI vendor. We’ve assembled a white paper based on our experience helping customers succeed with AI. To secure a copy, follow this link.
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.