How to recognize (and avoid) ‘AI washing’

With all the upsides resulting from the trend toward using powerful datasets to generate new value, come some clear downsides

artificial intelligence / machine learning / worker replacement

Artificial intelligence (AI) is permeating enterprise technology at a faster rate than ever before, yet it is still just the beginning of the adoption curve. From self-driving cars to intelligent assistants to cybersecurity to creating more personalized search results and recommendations, AI and its subsets, machine learning and deep-learning technology, are already making an impact on our experiences at home and at work.

However, with all the upsides resulting from the trend toward using powerful datasets to generate new value, come downsides. I’m not referring to the doomsaying that AI will destroy us, but the unfortunate fact that software and technology vendors looking to capitalize on the hype will exaggerate their AI capabilities to get your attention and boost sales.

The concept of “washing” isn’t new—we experienced it with “green washing” and even “cloud washing” and rightly should be skeptical of companies claiming to incorporate the technology into their solutions. There are obviously several key companies investing heavily in the technology and countless academics pushing the field forward. However, fewer than 39 percent of companies globally have an AI strategyaccording to a recent survey by MIT Sloan Management Group. The survey also revealed that only one in 20 companies have incorporated AI into their offerings or processes in any significant way.

These stats should create a buyer-beware situation and to reduce the risk of falling prey to these types of misrepresentations, companies should look at three things when evaluating whether a software or technology vendor has true AI capabilities.

The right employee expertise

First, and most important, make sure the vendor has the expertise and employees that have extensive education and backgrounds in AI and deep learning. This includes a wide bench of data scientists, mathematicians, architects and engineers. These people are responsible for coming up with the models and teaching the machines to understand myriad situations, plan future actions, predict their impact and learn from the results. It’s also important that they have backgrounds in visualization—which is the process of actually making sense of the data in a way that translates easily. Visualization is often as critical as the algorithms themselves.

Big, deep-pocketed companies like Google, Facebook, Microsoft, and Amazon are able to recruit or buy the best talent out there. Facebook alone has 100 employees dedicated to AI research, and they all have thousands of data scientists on staff, many of which are PhDs and experts from universities. So be sure to get specific about the staff behind the tech and their previous experience.

Data, data, data, and more data

The next critical piece of the AI puzzle is data … tons of it. When it comes to AI applications, the more data the better. Look at the scope of the vendor’s data and make sure they’re collecting information from a variety of sources. Ask them about the diversity of their customer base in terms of vertical industries, geographies served and sector sizes.

Again, not every company can have the scale of Google or Facebook, but the more depth and breadth of data will ensure there’s enough to correlate, find interesting insights from and learn from to make more valuable decisions.

A flexible, yet secure, infrastructure 

The last element is ensuring IT vendors have is the right infrastructure. Born-in-the-cloud companies understand the challenges associated with managing datacenters, governing access and securing the data. Another crucial role these companies need is devops: experienced teams that can stitch together the various technologies—such as Apache Spark, Hadoop, and open source components—needed to make an AI system work, as well as granting the right access to data scientists.

If your vendors are operating in the EU or serving customers there, they’ll also benefit from anonymization engines to maintain compliance with data protection regulations, which can make system development more difficult, or risk hefty fines and potentially worse. Indeed, companies everywhere could benefit from the added security.

AI is certainly where our future is headed. But be careful, the hype can be dangerous. If the term is misused, it will be at risk of becoming another broadly ignored marketing term. Software and technology vendors should listen to what Gartner recently recommended: “Use the term ‘AI’ wisely in your sales and marketing materials,” and “be clear what differentiates your AI offering and what problem it solves.”

Copyright © 2017 IDG Communications, Inc.

How to choose a low-code development platform