Three questions to ask before buying an AI solution
There currently isn’t any ruling authority that defines what vendors mean when they say they offer AI solutions.
As you go through the procurement process for an AI-based solution, be aware that providers are focused more on selling their solution than what the exact definition of AI is— and they may not even know how to define it.
This means you’re going to have to be your own advocate as you make buying decisions that will affect your whole organization.
But how do you ensure that when engaging with an AI vendor, you’re not jumping headfirst into the wrong pond?
Though it’s not an exact science, there are three key questions you can ask to bring clarity to the buying decision. And if any prospect can’t answer these, you’ll want to look elsewhere.
1. What qualifies your solution as AI?
This question may seem so basic that you don’t want to ask it— perhaps you feel like it would be offensive to ask, or rude to the representative you’re dealing with.
But it’s important to hear how the vendor answers. Plenty of providers out there are touting things as AI that actually aren’t. For instance, some vendors refer to basic scripted automation as AI, but this automated process is not being required to think or act like a human. It doesn’t learn as it goes, either.
How does the solution leverage AI, and what does it actually do? Make sure the person pitching you can truly explain it. Sometimes, the simplest questions are the most important and the most difficult to answer.
2. How are ethics questions being addressed?
Because AI holds such powerful potential, opportunities for its use are vast, whether it’s using the data to manipulate elections or people’s purchasing decisions– and essentially every other nefarious application you can think of.
Though capable of finding connections and computing problems that humans can’t, AI lacks common sense and the ability to reason.
At worst, these deficiencies could lead to ominous scenarios like those from War Games and The Terminator. Less catastrophic but still undesirable are AI’s demonstrated bias and questionable uses. Does the smart TV in your office, for instance, really need to eavesdrop on everything you say?
With these issues in mind, a vendor needs to have a well-thought-out policy regarding
the ethics of AI, one that has integrated this consideration into the ways they train their employees and their machines.
A formal policy that dictates ethics and some sort of ethics training for staff is important.
3. What do you do about the cold-start problem?
There is a potential problem in the computer-based information systems that involve a degree of automated data modeling. Specifically, the system cannot draw any inferences for users or items about which it has not yet gathered sufficient information. This is called the cold-start problem.
Vendors don’t just push a magic button and turn AI on. Instead, AI “grows smarter” by gaining intelligence in the form of data as its use progresses. But as a company, you need to get started as quickly as possible– you don’t have time to wait as the solution collects your company’s data to finally be able to start using it.
It’s critical to overcome the cold-start problem, so the vendor you work with needs to have a solution for this. If they don’t, that’s a red flag.
Get what you need
As if the concept of AI itself wasn’t difficult, it has become more so as you search for an AI-powered solution for your organization.
Because there is no single standard governing what “AI” means, vendors are free to claim that their solution has AI capabilities. They may actually believe it does when in reality, it’s just scripted automation.
You will also need to know how vendors will train the AI and how quickly it will return value, as well as address ethics issues. The above three questions will help you find the AI solution that is actually AI-based and meets your specific needs.
This article was contributed by Dr. Akhil Sahai, chief product officer, Symphony SummitAI.
18 July 2019