Using AI To Improve Your Predictive Capabilities

Earlier this year, a new report from the MIT Sloan Management Review highlighted the poor progress being made with digital transformation efforts across industries. This is especially so in terms of Artificial Intelligence, with few companies progressing beyond pilot projects that have done little to truly transform the way companies work.

At the recent EmTech event, Carnegie Mellon's Zachary Lipton argued that half of the problem is that the very term Artificial Intelligence has suffered from excessive hype.

"It's getting harder and harder to distinguish what's a real advance and what is snake oil," he said.

The Truth About Predictions

Human beings can earn themselves a significant amount of money if it turns out that they are particularly good at making predictions. We know this to be the case because there are many people who make a living speculating in the stock market, etc. 

However, humans are not very good at making accurate predictions, and that is why we turn to machines to help us out with this. Using machines allows us to make more informed and more useful predictions at the end of the day. A huge amount of data can flow into the technologies that we use to make predictions, and this will allow us to craft our predictions as accurately as possible for maximum impact.

Realistic Origins

The University of Toronto's Ajay Agrawal, Joshua Gans, and Avi Goldfarb argue that the most realistic beginning for AI in organizations today is in lowering the price of making predictions. In their recent book, Prediction Machines, they suggest that by focusing on making predictions easier, they not only make it more likely that AI projects will succeed, but they will also lead to substantial gains for businesses.

Whilst they highlight the "predictive purchasing" patent registered by Amazon that would potentially see the company using their extreme understanding of each customer to buy things on their behalf (with free returns then offered if it wasn't what they wanted), a more realistic application is in the continuous experimentation Amazon do to improve the customer experience offered to users.

It is, perhaps, no surprise, therefore, that luring Angela Pesta from Amazon was one of the first steps taken by Dana Dunne, CEO of travel giant eDreams when he joined the company in 2015. At the time, the digital tools offered by the company were widely regarded as extremely sub-standard, and getting the digital side of things up to scratch was a major strategic priority.

A Burning Platform

Dunne had a clear mandate for change as the company was fresh from a catastrophic stock market performance that had seen shares plunge 60% within months of the company floating on the stock market in April 2014. The plunge wiped over 1 billion euros from the company's market value.

The company, which despite perhaps not being a well-known brand, is the largest retailer of flights in Europe and serves nearly 40% of customers via mobile. Travel is a fundamentally unique experience, and the digital strategy was driven by a desire to first better understand customers, and then deliver a more customized experience to each customer.

"Technology can help customers much more and create a much more personalized experience for them," Dunne told me recently. "We wanted to leverage technology to set us apart and really help the traveler and help us to develop a real competitive advantage."

A recent study from Google found nearly half of us are very comfortable conducting all aspects of the travel experience via our smartphones, whether that's researching, planning, booking or reviewing our travel. What's more, around 70% of travelers always have their mobile with them when they're traveling. It's clear that mobile is crucial for any travel company, and this was the obvious first step for Dunne when he joined the company.

The early improvements in the mobile app were then built upon to create a detailed understanding of what the customer wants and how they go about booking their travel. The company utilizes everything from virtual reality to eye-tracking software to better understand the customer, alongside extensive quantitative analysis of user behavior via the app and web interfaces.

This has culminated in a recent study conducted by the company, called The Meaningful Journey, in which they studied several thousand people from across Europe via both surveys and real-life analysis from the company's user experience laboratory. The research aimed to truly understand what travelers want across Europe and to use this to inform the customer experience.

The Skills to Succeed

A recent survey by EY highlighted a lack of technical skills as a key barrier to moving AI-driven projects out of the pilot stage and into something more substantial. It's a challenge eDreams have also had to face, but while they have grown their developer base to around 400, they have also attempted to involve all aspects of the business in this journey. Whether it's the people who run the user experience lab, or data analysts who help to codify the input from the customer.

"We really bring all of these multi-functional teams together, and so this consists of hundreds and hundreds of people," Dunne says.

As for the MIT report, I opened this piece. However, while the technical skills gap is perhaps the most well-publicized, the skills gap in the boardroom is more important. Many of the companies identified by the report as leaders in digital transformation had leaders who got the new ways of working made possible by the latest digital technologies.

This is often especially challenging when attempting to build an experiment-driven culture that inevitably leads one to accept the limitations of your knowledge and to let the data from experiments drive your decisions. Many companies have grown up with a culture whereby leaders are expected to have the answers, so having that humility can be hard.

At eDreams, they're making several thousand updates to their digital platforms every year as a result of the ongoing experimentation they're running, so it was crucial that their leadership team bought into this.

"The board has been extremely supportive of the business," Dunne says. "They ask good questions about it and know that, at the end of the day, it's down to the management to run the company. We've actively selected leaders based upon their willingness to learn, and our leaders have bought into this."

Using AI and data in this way may not be as sexy as some of the more breathless predictions for the technology suggest, but creating an organizational "prediction machine" is not a bad place to start your digital journey.