How to Guarantee the Success of AI at Your Business
History has, and always will, repeat itself. So if we optimize based only our own history, won’t we merely repeat ourselves…faster?
Across all of the conferences I’ve attended in 2019, there have been many, many discussions surrounding machine learning (ML) and artificial intelligence (AI) and their application to optimizing data. In every discussion, someone asks the question about the level of effort of the “machine” to understand the data. And in every answer, you hear the same thing: It takes time and resources to optimize. Many articles and studies confirm that it takes years and a ton of resources to feed the “machine” with information, so that it can learn your historical data.
Optimized historical data? Is that really what we want? Any analyst will tell you that while optimizing your own data can drive some efficiencies, the optimized point will only be as large as your best historical performance. In a world where better performance is the key to increasing revenue, it’s necessary to use data points that you don’t have historically.
For example, if you want to reach consumers for sales, customer support, collections, et cetera, and your data suggests that you always reach the most people at 3 p.m., then your optimized data would lead you to call everyone at 3 p.m. But what external data will tell you is that you have multiple time periods in a given day in which individuals are most likely to be reached. Those individuals may be grouped together in clusters, but not every one of those individuals is available to be contacted at 3 p.m. Think about that logically for a second: Assuming a random consumer population, do we really think that everyone in the country can be reached at 3 p.m.? Of course not!
As you spend money to create and build an AI or ML tool, you will inevitably need to get your hands on external data on top of your historical data. That external data introduces hidden elements: individual consumer characteristics that make each consumer unique. Said another way, if you want to actually contact someone, don’t call them at 3 p.m. when external data will tell you that they are most likely to answer the phone at 10 a.m.
I’ve used time of day as an example, but just imagine all of the other data points used to reach consumers: phone numbers, addresses, email addresses, et cetera. In an historical data world, the data typically comes from only one place, and the optimization will never know if that information is bad or incomplete. But with authoritative external data, the optimization tool now has a means to arrive at better results.
Does external data cost money? Of course. Being better means thinking outside your history and incorporating intelligence that actually drives improvement. Don’t continue to do the same thing over and over to get to the same end faster. Get to a better, different end—faster.
I’m not suggesting that you should not embark on machine learning or artificial intelligence. They are the wave of the future. What I am suggesting is that if you do embark on this journey, you must use information you don’t have today, or your investment will lead you to the same point you’re already at. Don’t do half the job; do the whole thing and reap the benefits and rewards.
Insanity is doing the same thing over and over and expecting different results. Whether you are currently on the ML and AI path or not, investing in external data can only mean better performance. And we all know that with increased performance comes increased revenue and profits!