Natural Intelligence and Artificial Expectations

Natural Intelligence and Artificial Expectations

By: Gael DE Talhouet, Vice President Brand Building, Essity GMBH

Gael DE Talhouet, Vice President Brand Building, Essity GMBH

Contrarily to most legacy IT solutions and platforms, AI is not a tech tool that runs in the background. It is not an enabler that business and customer-facing people can ignore. As a well-known quantum physics principle explains it, the very fact of observing changes the experiment. What I mean with this, is that AI does not work with Euclidian, linear rules. It is not like you have a business problem, then you brief the first techie you can find, then he/she goes back to the cellar and then delivers something that works six months later, that you could not care less, as long as it works.

The reasons for this are multiple.

One, most business owners do not have A problem. They want to sell more, sell to more customers, and sell with higher margins, none of them being a defined problem. I am not talking here about production and supply chain which are trained the organization to work on KPIs and can define a problem, quantify it, and expect a solution.

This is the second reason. Business, sales, marketing organizations can not envision what the solution is, even less what the outcome would look like. Like in quantum physics, most people cannot represent what a more than four dimensions world looks like, nor antimatter. Using SAP to invoice my customers is something I can think of, using AI to sell more is something I have no clue about.

"AI transported our team into a quantum world, where energy is stored and released, creating paradox, where matter (campaigns) interact with energy (money) and time in an “unnatural” way"

The third reason is that, for business owners to start having a clue, it needs interacting with the solution. Just like “I can transport your voice through space wherever you are” has no tangible reality vs. using an iPhone, which is concrete and self-creates the need to solve. The implication is that no analyst in the world can transform AI for Business into requirements, because it will be the user interacting with the solution that will create the need.

The fourth reason is the illusion of clean data. I have seen teams and organizations working for years in a holy quest to clean the world of data. Four years later, they are still, in their own words “fixing the basics.” This does not even take into account the fact that in between, legislations will have changed (etc. GDPR), key players will have changed their rules to data access and formats (ex. Facebook), new players will have emerged, etc. Therefore, the only way to start figuring out what on earth could AI do for business is to begin, and start with “dirty data.”

The fifth reason is the AI Business Case. The sacred “Business Case,” the magic that every organization entertains, according to which creating a few Excel sheets and PowerPoint slides, just like religions write their sacred verses, will uphold the truth, prevent the project from failing, and lead the believers to paradise. This is another opportunity to lose time and resources, thinking and talking, while any measurable business output would again only come from using the solution to define a problem, which once solved, can deliver a business impact. Like most successful start-ups have found their business model along the way (e.g., Google, Facebook), AI will define its business opportunity along the way. No one has ever asked for a solution to “make the world more open and connected” nor can you build a business case out of it. People have found an interest high enough in sharing pictures of their cats to swallow a massive amount of undesired advertising, and finally unlock business potential.

The AI “Business Case Fun” has another part to it. Professor Erik Brynjolfsson, Director of the MIT Initiative on the Digital Economy, has put a name on it: the “Productivity J-Curve.”The principle is rather simple: new technologies require additional investments and time, classically not measured, which trigger a significant loss of productivity at the beginning, before raising again, hence a J-shaped curve. Think of team adoption, ways of working adaptation, change management, learning how to use, creating ownership, trials and errors, explanation of output vs. expectations, integration into current business practice. The details for this fascinating research paper are in the footnotes of this article.

This brilliant J-Curve Principle shed light on one of the experiments we made of using AI to optimize digital campaigns. More than the expected first few weeks of “calibration,” we found the team struggling with manipulating the object for months. Even after the first weeks, nothing made sense. Because nothing in AI was like a “normal” business. Nothing in AI fitted our natural intelligence and capability to anticipate. Nothing in AI was linear, nothing in AI fitted to “classical mechanics” of marketing. Normally, “naturally”, business people develop a sense of consequences and anticipation. This is what is called “experience” and “natural intelligence”. The capability to imagine / anticipate if the impact of one action will be big or small. Simple example: changing the small type font on a packaging, impact small. Increasing your prices by 50%, impact big. More digital example: changing one word on a search ad copy, impact small. Replacing paid search ads by a CRM emailing campaign, impact big.

In using AI for marketing and sales, the team was confronted by a world which did not obey the law of classical marketing mechanics. A tiny change had a massive impact on the AI, apparently going back to infancy and re-starting from scratch, while significant changes were “swallowed” by the robot-like nothing happened and impact was observed almost immediately. For months, the team struggled with a defiant law of classical mechanics: no cartesian causality, no more “Natural Intelligence” in anticipating big change vs. small change, no more time quantification associated with events in an understandable way. AI transported our team into a quantum world, where energy is stored and released, creating paradox, where matter (campaigns) interact with energy (money) and time in an “unnatural” way.

In the end, we got what we wanted: lower Cost Per Acquisition and more Sales. But the “business case” approach, balancing investment in money, resource, and time were defied by the two things we learned the hard way: the J-Curve of productivity and the Quantum Mechanics of AI.

In conclusion, we call to all business leaders. AI is “artificial” in the way that it is “non-natural” for companies. AI cannot by an IT project; AI cannot be planned (by the way, AI does not live in a plan for its non-Euclidian nature), AI cannot be business planned, because only the interaction between the user and the solution will reveal if and how it has business value.

Check This Out: Top Retail Artificial Intelligence Companies in Europe

Weekly Brief

Top 10 Artificial Intelligence Solution Companies in Europe - 2019

Read Also

Thinking out-of-the-box to launch business beyond borders

Thinking out-of-the-box to launch business beyond borders

Bill Bass, Chief Marketing Officer and Kara Lawson, Senior Director Ecommerce, FULLBEAUTY Brands
Rebuilding the Ecommerce Value Chain

Rebuilding the Ecommerce Value Chain

Jackson Fernandez, Head of Ecommerce, PUMA Group [ETR: PUM]
The COVID Catalyst and The Paradigm Shift

The COVID Catalyst and The Paradigm Shift

Alexandra Hourani, NA Digital GMM Director, Converse