The WHY IN AI

Working Paper No. 28

The Why in AI

Artificial intelligence more intelligently in marketing

By Björn Bringmann & Gordon Euchler

AI plays a decisive role in many areas today; in marketing though, there is still a need to catch up. Deloitte experts show the reservations and opportunities.

AI is one of the most important business topics today. According to a Deloitte study, 93% of business leaders see AI as key to staying ahead of the competition - today and in the future. AI is on everyone's lips. And it's always a new topic of conversation.

Every year AI takes over some new human tasks.

Since its beginnings in 1956, AI has already come a long way, including lows - the so-called "AI Winter". Yet every year new, previously unimaginable capabilities are added. In 1997, the AI Deepblue beat the world chess champion for the first time. In 2002, the world saw the first intelligent robot vacuum cleaner. And the first autonomous driving car.

All these inventions are absolutely groundbreaking when they see the light of day. But day by day, we get used to them and after a couple of years they no longer seem exciting. Or even AI. But merely a “normal computer”.

But why shouldn't this timeline continue to run like a line toward the future, with AI acquiring a few more human capabilities every year? Because that's not Artificial Intelligence, but - if you'll pardon the expression - it would be a bit stupid.

From “imitating human behavior”, to “daring to try something new”

All the AI innovation mostly mimic human behavior. The car that drives like a human. Just a little safer. Or the chess-playing AI. And the dust-swirling vacuum robot. But is imitating human behaviour really all we want to achieve with AI?

When humans fulfilled the other great dream, that of flying, they also first imitated birds. They learned to fly. Further and further, faster and faster. But how much do today's jets still resemble birds? Not very much. Doesn't a similar approach make more sense for AI?

Using AI creatively

So the question is: how can AI be used today in such a way that it doesn't simply copy human behavior patterns? How can it, instead, breaking new ground.

Daniel Könnecke, partner at Deloitte, says: "A different approach to AI is needed. Less copying of human output. But more understanding HOW AI thinks.  It's about how AI thinks differently than humans and what you can achieve differently with it.”

The often overlooked question: How does AI think?

In classical programming, you define the input, apply the "transformation rules" to it, and get the output. This means that you have to define exactly these transformation rules. For example, how to determine the accounting-profit, how to calculate the expected sales in the case of a price increase or, in increasingly complicated multi-level models, how to model the purchasing decisions of a person. Then - and only then - can a computer calculate the output. Faster than any human but always limited to exactly these "transformation rules".

Machine learning of an AI works the other way around. Here, the system is fed with huge amounts of data - you give the AI as many related inputs and outputs as possible. AI then takes over the task that humans had in classical software development. From the input/output examples, AI works out the transformation rules (i.e., the program) that lead to the output. Thus, not the result, but the regularities.

So if AI is not fed rules, but works them out, then you have to use that strength.

Because someone who works out and discovers the rules themselves - more than humans would manage - is particularly good at two things. AI can keep discovering new rules to tackle exactly a single problem. In other words, it can solve an extremely narrow task extremely well. And AI can repeatedly discover new possible solutions for a very open task - in other words, it can work extremely broadly. That creates opportunities in marketing.

Brand management - taking a KPI to the absolute extreme

So if there is only one task, an AI can process much more complex input and generate "transformation rules". Rules that no human would come up with and rules that still achieve better results. And that's exactly the task in brand management.

There is a defined input for almost all brands. The target groups. The brand dimensions - from luxury to sportiness or even psychological dimensions such as dominance, adventure and the like. And the output dimensions are also defined: namely, the entire brand presence, which is supposed to do justice to these dimensions. So far, one follows the ratio of input to output as well as one can. No matter how good that is: AI is predestined to do more here.

Florian Klein, Director at Deloitte adds: "AI can read and analyze all brand assets and all touchpoints available on the internet about the brand - whether paid, owned or earned. In fact, AI can analyze the entire UX including consumer interactions and their impact on positioning dimensions. So AI can holistically tell to what extent the brand in its entire appearance meets these dimensions and where it doesn't." This makes much more comprehensive brand management possible, and the first approaches to this are already working, such as Gnosis from Deloitte's innovation team. An exciting approach: because in this way we are approaching 100% brand control for the first time in the history of marketing.

Fuelling Creativity - Discovering New Paths

Another exciting marketing opportunity for AI is that it produces many more solutions across the board. Solutions that a human being would not have come up with. Although these often seem nonsensical, that is precisely what can sometimes be useful.

So if you give the AI a task that is simple for a human, it sometimes performs surprisingly poorly. You plan a small dinner party. Dinner is to be served in the living room. Unfortunately, the dining table is wider than the door. To get the dining table into the living room, you have to ...

The AI does not detect the obviously sensible rational simple answers: "You have to turn the table to the side so that it fits through the door." Or, "You have to take the table legs off." But it does discover seemingly nonsensical answers like, "Take the door off its hinges. You have a saw - saw the door in half and remove the top half."

Or she has to solve a jump&run game as best as possible. What does it do? She gives the character longer and longer legs. Longer and longer. Until it no longer has to learn to walk, but simply falls over. And into the goal.

Both answers are solutions only to a limited extent. But both answers suggest ideas that the authors would probably not have come up with themselves. In short, AI does not produce the perfect solution, but it does produce surprising suggestions. This in turn often triggers a new and this time really convincing flash of inspiration in creative people. And unseen solutions are the ones to look for in innovation and creative departments. An exciting possible application of AI in the future. Not replacing people, but firing them up creatively Matin Ebrahimchel, partner at Deloitte says: "I clearly plan to use AI as a stimulus-giver in creative workshops of the future - it simply expands our creative radar." It is also conceivable to use AI for campaigns with a very narrow creative concept - such as Haribo's children's voices campaign: then AI can produce more and more new variants here as well.

In short, we believe that AI can do more than replace human activities. Rather, it can break new ground. This is of particular benefit in marketing. We are convinced that there is more to come. And we are already excited to discover more of it.