Big companies inspired by AI start-ups have been fighting over data scientists in the hopes of finding that little nugget of gold that will optimize and automate their operations, identify which customers will bring in the most revenue, or even lessen the toll of repetitive tasks that bring no added value.
Like many other inventions before it, artificial intelligence is a technology that should be subject to the same cycle of adoption, according to Gartner’s now-celebrated hype cycle.
In my opinion, we’re currently scaling the upward slope of inflated expectations. All the symptoms of hype are present, including the following:
Inevitably, this overblown enthusiasm will lead to disappointment. We’ve already seen this type of hype cycle play out in several areas, notably regarding the web over 20 years ago. The same thing happened with virtual reality a few years ago. (Actually I think we’re still experiencing the trough of disillusionment with regard to that particular technology.)
The disillusionment phase obviously brings its share of negative consequences. Projects, programs, and investments are suspended, jobs are lost, and dreams are broken. One could also take the view that these readjustments of the market represent an appropriate return to an equilibrium state where expectations and delivery are more optimally aligned.
In my view, this disillusionment is best expressed by the idea of a group of business managers and researchers who are trying to work together. The managers are looking for a breakthrough concept that will lead to guaranteed profits within a predetermined length of time—generally in the short term. The researchers, however, feel a lot more comfortable with longer term work cycles and indeterminate results, and naturally do not share the same gauge of success. In short, the researchers formulate their hypotheses and see value just as much when those hypotheses are confirmed as when they are negated, as opposed to the managers, who, for a variety of reasons, don’t have a lot of interest in scientifically proving that a particular idea is faulty.
Disappointment will come particularly when these two different realities collide, but also from the process that allows them to get a taste of the advantages of AI. Conferences and discussions surrounding AI focus generally on analysis and data activation, or in other words on insights. But there are two essential phases you need to deploy before you can get there: data capture and engineering. According to three experts invited to a panel discussion on the subject, these two phases represent between 80 and 90 percent of the work in AI.
What we see a lot in digital is that data are basically accessible, but most companies capture data in a very approximate way. The situation is even worse when it comes to engineering, which involves organizing, structuring, cleaning up, and preparing data so that the work everyone wants to accomplish can actually begin. Artificial intelligence will only take off if those managing its emergence accept and trust the process.
People developing AI business use cases need to consider this work as part of their costs before vaunting the benefits of AI if they want to grade the downward slope towards the trough of disillusionment.
Consider this picture: There’s a desert you want to cross. Only a few will be successful. Do you trust the people lugging hundreds of litres of water along with them, or those who find a way to create water as they travel? The first group won’t make it to the other side, since they’ll be weighed down by their heavy baggage. These are the companies that hired an AI team too quickly without having a sensible, adapted business plan.
Concentrate a large part of your work on the capture and organization of data to make them exploitable and reliable. Well-structured data are the oxygen for every AI initiative. Take advantage of the upward slope of the hype cycle, when budgets are larger, to accomplish tasks that will quickly become unpopular when things are sliding down the slope towards the valley of disillusionment.
Taking time to properly define the problem will save you time in the end. While it might seem like a good idea to apply emerging technologies or models (chatbots! QR codes! multitouch attribution! influencers!) that may appear to resolve the issue, it’s crucial to ensure that stakeholders have a deep understanding of the problem and have a broad perspective on possible solutions. This is where creativity comes into play, but it cannot thrive without this basic foundation. Who knows, you might even come to the conclusion that you don’t need AI to solve your problem or accelerate your growth!
Give your new multidisciplinary teams the freedom to make mistakes since they’re exploring uncertain territory, collaborating together for the first time, and may not always know where their research will take them. For example, don’t take for granted that a mathematician is comfortable with basic marketing principles. The important thing is to establish a culture in which people learn from their mistakes and that encourages the various teams to explore the realities of other involved teams so they can reach a level of mutual understanding that will lead to exponential benefits over time.
Finally, automate the little things and start with simple algorithms that create small advantages. This approach allows you to begin the collaboration process between your researchers (mathematicians, data scientists, etc.) and businesspeople. This step is more than just an end in itself—it’s a way to establish the groundwork for these groups’ crucial collaboration and manage the expectations of those involved. It will also have the effect of decentralizing risk over many small ideas instead of concentrating everything on one big idea whose outcome is too far in the future and too costly. In short, if your team can’t build a cottage in two weeks, there’s a good chance it will never be able to build a castle in twenty years.
Our authority isn’t built on basic AI research. Rather it’s our expertise and experience with dozens of different industries and business models that enable us to apply that learning to artificial intelligence in a marketing context. We aim to participate in our clients’ efforts in a concrete way by supporting them during their preliminary work with data, but also through helping them deploy AI use cases that are low risk yet have promising expected returns.
To accomplish this, our data science and technology team applies AI approaches to marketing through dozens of algorithmic models that have demonstrated applications to business issues. For example:
When used together, these simple, effective models can also become the basis for a more complex approach that will lead to major differentiation for our clients. But they also have great intrinsic value.
Our digital transformation team supports a range of companies in establishing the aforementioned steps, in creating and organizing the internal teams affected by the advent of artificial intelligence, as well as the marketing and creation of new business models now made possible by AI.