But what exactly are they ready for?
Spurred by the excitement around IBM Watson and Star Wars commercials, tech giants like Adobe and Salesforce are capitalizing on a lack of clarity around AI, a murkiness left unexamined in the media's eagerness to create hype (and founders' eagerness to scare up venture dollars). However, businesses need to start separating hype from reality and start asking questions about the intelligence of the solutions flooding the market.
The following are three questions that will help organizations properly assess the value of any AI solution -- before and after adding it to the budget.
1. Are these "AI technologies" actually new?
A few months ago, Salesforce unveiled Einstein, the AI capability that's supposed to make their platform better at identifying valuable customers in their sales funnel. But recommendation or next-best-action engines have been identifying high value clients for years. What exactly is the new AI function here, "Einstein"? Is it just a better algorithm, or is the idea that it can learn to properly identify a valuable customer without human assistance? The latter would be impressive, but I'm skeptical that it's possible. Einstein also claims to rank leads by value and suggest when to approach them, but any programmer with a basic understanding of machine learning can design a tool to do that.
Adobe launched Sensei as "a unified AI and machine learning framework," claiming that it will help businesses find new "look-alike" audiences, but that's become standard issue on most good data management platforms. Sensei also supposedly has the ability to suggest messages that will resonate with a customer. However, I know from deep, personal and sometimes painful experience that this in fact takes years of investment and thousands of experiments with millions of samples. I question how long Sensei has been studying this subject, and how far along the learning curve it is.
2. Have these AI capabilities been developed with a specific business case in mind?
Currently, there are no "plug-and-play" broad AI solutions that can add value right out of the box. Machine learning algorithms only improve by spending months, if not years, digesting data to refine the models and serve specific business needs.
As one example, IBM's Watson API is impressive and has already been tackling questions around weather reporting and cancer diagnosis. But a ton of data have been fed into Watson to develop these capabilities. Something like Watson cannot be blindly applied to any new problem without the requisite contextual learning beforehand. If, for example, I take Watson's horizontal API and apply it to, say, marketing messages, I'll get little-to-no actionable insights. More than that, I'll likely get the wrong insights. Watson's API qualifies "Attention please. Our offer ends today." as "sadness." But does that sound sad to you?
3. How do you know that your AI solution is better than what you already had?
This Harvard Business Review piece notes that while machine learning can help generate unique value insights, it's also true that "they'll often fail miserably if you try to apply them to something new, or, worse, they may degrade invisibly as your business and data change." It's imperative to test the efficacy of the AI recommendations against your current systems to ensure your solution is working properly, and not giving you potentially damaging results! This requires organization and process changes that few businesses realize they need to make in order to ensure machine-learning technologies are benefitting their operations.
Unless you're actively measuring results and holding the machine-learning models accountable, you can never be sure of your ROI. Furthermore, considering how many of these "AI solutions" are just repackaged existing technology, it's even more important to be aware of their outputs and keep track of what they're telling you and the value they bring.