AI has long been hailed as the crowning a achievement of technological prowess, but the big question for businesses is “Are we ready for AI?” and more importantly “Is AI ready to actually be of service to us?”. There are already many service providers offering impressive new AI capacities to their clients, but what is really being sold.
“AI” is a very broad term that is certainly more prevalent in the SciFi category than in the realm of practical solutions. When we witness the wonders of IBMs Watson and the thrills of the latest Blade Runner flick in the same day, we wake up the next with a fuzzier understanding of what this phenomenon is capable of.
So before running off to purchase an AI solution to whatever needs your company may have, the savvy business asks the following important questions.
1. How new is this AI?
When Salesforce announced the arrival of “Einstein”, they declared that this AI solution was capable of enhancing the Salesforce Platform’s capacity to identify the best customers in a sales funnel. While this sounds impressive, it is important to remember that this has already been accomplished by recommendation and next-best action engines for a while now. So what is it that “Einstein” actually brings to the table? Does it provide high-value customer identification with no input from human users? Is it just another sort of algorithm? The first option would be highly impressive and represent advanced machine-learning, to the point that I doubt it’s possible (but maybe?)
This AI solution also claims it can rank leads by value and make recommendations for the best time to contact them, but this is also a fairly simple trick that could be performed by an amateur programmer.
In a similar way, Adobe invested heavily in marketing their Sensei program as a” framework for machine learning and a Unified AI”. They claimed that Sensei would d allow them to clearly identify “Look-Alike “audiences, but isn’t this a feature that most data management systems offer as a standard equipment? Then, they also claim that Sensei will make recommendations on the best things to say resonate with a certain platform. This oneself is quite a feat that takes many years of learning and studying to achieve effectively. There is much experimentation and contextual learning involved and this demands considerable investment. The big question here would be “How much learning and studying has the Sensei AI solution actually accomplished? “
2. Have AI been developed to address a single aspect of marketing?
As yet, there is no “Plug’n’Play” solutions to AI for marketing or business. The fact is that machines have to spend many years ingesting data and constructing business models before any of the productions will be of any use.
Let’s take a look at the impressive API from IBM Watson. It is the most impressive AI of its kind and has handled all types of questions from predicting weather patterns and analyzing cancer diagnosis and even the best strategies for winning at Jeopardy! Of course, none of this precise calculation was an overnight production. Watson needed indescribable amounts of data to begin to formulate these. Watson can’t be directed at a problem and asked to find a solution unless it has been fed enough contextual data to process the problem effectively.
If you were to Watson and apply it to creating marketing messages, it could get most of the meanings very wrong. For example, a message like “Attention please. Our offer ends today” could qualify as a very sad message, but is that really the case?
3. How do you know the AI solution improves on what was already there?
An Article in the Harvard Business Review noted that, even though there can be some unique and valuable insights from machine learning, they can “fail if applied to something even slightly unfamiliar or new”, they may also begin to degrade as your business and data change over time. Look at this post from Harnham on the matter.
Unless you are watching the AI carefully and monitoring all of the results it is almost impossible to predict the type of ROIS you can really expect. Full awareness of the outputs you are expecting is essential to ensuring you get what you really want from the techno support.
