Commentary

What You Need To Know About AI, But No One Will Tell You

Artificial Intelligence is a game-changing technology. It will have more impact on our lives and on business than the Internet did in the 1990s or mobile phones did in the 2010s. We are in the early stages of this revolution, which means we are all learning. Hopefully these tips can help demystify the hype and help prepare your business and your brand for an AI future.

The best AI is being done on free software. There are a thousand companies selling AI solutions, with thousands more coming. When navigating those waters, it is helpful to know that no one has a monopoly on AI. You can run world-class AI algorithms for free, as long as you have the right personnel and training. Kaggle is a data science competition website, which is essentially the major leagues for analysts. There are real data and real stakes, and those competitions are always won using R or Python, both of which you could download for free.

This is great news. If Google, Amazon or Microsoft had a monopoly on AI, innovation would be much slower and expensive. As marketers, it is empowering to know that you can apply world-class AI for your brand, as long as you have the personnel who can tap into it.

Sometimes, AI isn’t as good as college-level statistics. The increased hype around AI has led companies to try to use it in every situation. What no one will tell you is that standard methods such as regression modeling are often within a few percentage points of performance of a machine learning algorithm. Sometimes simple regression even wins. Of course, it depends on the situation. A regression model would never beat AI in image recognition, for example. It also is important to know if those few percentage points between machine learning and regression matter or not. If a percentage point means millions in revenue, or an increased risk to employee safety, maybe it’s worth the extra trouble and processing time to do machine learning.

You probably don’t need deep learning today, and maybe never. Deep learning is another buzzword with a lot of momentum. It is an incredibly important technology that will lead to some of the most significant AI innovations, such as self-driving cars and voice recognition. However, most business cases today don’t deal with images or audio. For most, just trying to make an accurate forecast of next quarter’s sales or predicting a customer’s next likely product, deep learning isn’t necessary. Other machine learning techniques not only take less time, they are also more accurate on an average business data set.

The biggest barriers to AI are people, not technology. AI absolutely has the potential to transform companies and entire industries.  It can make your marketing and brand experience much more relevant and highly personal for your consumers.  But the race to an AI future in each industry will not be won by the company that can buy the biggest servers. Rather, the winners will be companies that have the most buy-in at the executive level.

Are you willing to reorganize your entire company around AI? How many degrees of separation are there between your CEO and your head of data sciences? When your AI forecasts a change in the market, do you trust it enough to change your prices and inventory? These are the types of people-related hurdles that will determine if companies and brands are truly ready for an AI revolution.

1 comment about "What You Need To Know About AI, But No One Will Tell You".
Check to receive email when comments are posted.
  1. James Smith from J. R. Smith Group, July 20, 2018 at 12:47 a.m.

    Jesse: Thanks for the great piece on AI.  I "get" your remarks about regression. Years ago we did some mulitvariate linear regression for a client to predict TV ratings from content analytic variables. Accuracy proved to be within 2/10ths of a rating point. Was it ever used.  Not that we know of.  Probable reason was lack of interest on part of senior execs, some of whom were very senior "creatives."

    To add a dimension to your points...it's not just getting C-level buy in for AI, it's getting them more educated on what, as you say, "college-level statistics" in the hands of a good analyst, has the power to do for their brands. If they don't have a "stats" mindset, then AI will become even more of a black-box.

Next story loading loading..