PORTO: While the data processing capabilities of artificial intelligence (AI) systems have garnered enormous coverage, brands are beginning to deploy AI on human emotion, delivering uplift for sportswear companies, airlines, and banks alike.

This topic was discussed at the recent I-COM summit in Porto, where a group of brand representatives shared perspectives on using artificial intelligence in the development of communications and the generation of messages to drive behaviour.

"What if we could decode language to evoke different emotions and move consumers to act?" asked Yannis Kotziagkiaouridis, global chief analytics officer at Wunderman. (For more details, read WARC's exclusive report: How AI is helping Under Armour, Air Canada and Barclays to drive behaviour.)

"Language can be organised algorithmically," he said. "You can look at not just the words that together form a meaning, but you also have the position of the language on the copy, you have the call to action, the description of the product, the emotion and the formatting".

"All those things are components of an algorithm that we have the ability to capture so we can test and see how consumers react," he added.

"In essence, if we know everything about someone, we can structure language using all these elements. We can expose an individual and read the results, and if we can do that at scale, rapidly, we can start making the connection that this individual responds to this type of language and emotion, and it gets this type of action."

Air Canada, for example, is using artificial intelligence to adapt its "Deals of the Week" emails so that the combination of copy, subject line and image triggers the emotions in readers most likely to lead to a booking.

Chantal Berthiaume, the airline's director of commercial analytics and customer insights, described how her team increased open rates of its "Deals of the Week" email by almost 50%.

Messages that included in the subject line the words, "Sitting down?" generated a degree of anxiety among readers, but also higher open rates.

"We would never write this," Berthiaume explained, "the machine picked it, and the machine was right. It did so much better than what we were writing".

Data sourced from WARC