Getting personal with AI
I had the pleasure of interviewing Chris Pook, VP Customer Retention from Lyst (the world’s biggest online fashion site) at the Email Insider Summit on the topic of Predictive Emails: How Marketers Can Bet on Future Customer Behavior. Essentially the conversation revolved around using data and AI to deliver tailored communication to customers.
Of course, AI is often described as being over hyped with one recent survey suggesting that almost 50% of marketers felt that AI was more fantasy than reality.
AI is nothing new. Did you know that the first AI workshop was held in 1956? But it was soon recognized that the development of AI would be hampered by the processing speeds of computers. And it’s only been the last 20 years where raw computer power has been able to start delivering the AI vision (Remember Moore’s law? Speed and memory capacity of computers doubles every two years).
And yet aren’t we kind of using AI anyway without knowing?
Hands up who has used the Google search bar, Siri or Google Translate in the last 24 hours? AI in action in all cases
There is nothing magical about AI: it’s just mathematics and applying statistics to predict outcomes.
Chris’s team uses data, models and AI not only to help understand what customers might want to see in their emails but also to predict what products are likely be discounted in the near future – one of the main reasons people use Lyst.
Lyst also very cleverly use AI to essentially make sense of all the product image feeds they get from thousands of websites. This allows them to clearly identify what the product is and how it should be described and tagged.
And that’s often the type of AI application that doesn’t get seen or talked about. I remember a couple of years ago literally asking an AI application to help me predict which customers were most likely to open a particular client’s monthly newsletter. Interestingly it showed that anybody who opened the first three emails we ever sent them or opened two consecutive ones since had a higher propensity to engage. Interesting because we were then able to re-imagine what the welcome programme might look like to deliver a higher proportion of those three email opens.
Of course AI is now encroaching on creative as well.
I read an interesting statistic the other day from Persado that in the UK last year the top 20 UK brands only used 293 different words in their email subject lines, and yet an average copywriter will have 40,000 words in mind. But a computer can literally have them all and as the associated AI comes equipped with NLP and machine learning algorithms, it can generate language in the brand voice and learn precisely what words and emotions drive the greatest engagement with customers.
So with AI is there a future without marketers? Luckily Chris and I agreed that even AI, for the foreseeable future, will need to humans to create hypotheses and to add ‘human understanding’ into the equation. Phew! We’ve got jobs for a while.
Thanks to Chris for sharing his experiences with us.