It was the kind of conversation about marketing with artificial intelligence that made SAS exec Monique Duquette tilt her head in confusion: a customer who said she wanted AI to tell her which campaigns were successful and which were not.
“I asked, ‘Do you have reporting?’ (She said) ‘Yes.’ Well, then why do you need AI?” the software giant’s national lead, customer experience and marketing told the 2017 Data Marketing Toronto conference on Monday.
For Duquette, AI can sometimes be a good example of something that fits the adage, “just because we can doesn’t mean we should.” Instead, she encouraged marketers to turn to AI only when it can do things humans can’t by themselves or with less sophisticated technologies. This includes sifting through vast data sets, identifying patterns, categorizing or quantifying information and automating processes at scale.
“We can seize AI for peaceful coexistence if we understand where we have human benefit and where we have the benefit to accelerate and interact with our customers without losing touch with the human side in those grand experiences,” she said.
Marketers also need to recognize that they can’t simply “add data and stir” and then expect to see a major boost in results, Duquette added. It’s a matter of teaching algorithms what it needs to learn.
Learning What AI Can Learn
In fact, she differentiated between AI implementations by the different levels of learning required by machines to understand and serve customers. Semi-supervised learning, for instance, might involve image recognition and classification, web page classification and speech recognition. Unsupervised learning, on the other hand, might involve market basket analysis at a retailer of what customers are putting in a shopping cart in order to determine ways to achieve an increase in spending.
One of the highest levels of learning is cognitive, Duquette said, where AI tools mimic human behaviour such as personal assistants, chatbots and Q&A systems. Even there, however, she said humans are still needed to shape the strategies behind them.
“(That kind of AI) is very good at natural language processing — looking at how we speak and interpreting it,” she said. “It still needs to get to the level of natural language understanding, when it becomes much more human-like.”
One audience member told Duquette she feels “cautious” about AI because, even if it has an ability to learn, it can make decisions that might be in error. This was something Duquette admitted is a common issue among many marketers.
“I thought (the main issue) would be the data, and it wasn’t,” she said. “Customers will tell us, ‘My only inhibitor is, I still feel like I need to explain why.’ In other words, why something happened.”
She pointed to an app created by the City of Boston, for instance, to report on potholes. Submissions poured in — but only from those who could afford a smartphone and take action through that channel. As marketers experiment further with AI, they’re going to need to be extra-careful about such biases, she said.
“Ultimately we’re still held accountable for the metrics and the strategies,” she said.
Data Marketing Toronto wraps up Tuesday.
Latest posts by Shane Schick (see all)
- Why are COOs the forgotten B2B marketing target persona? - November 19, 2019
- ActiveCampaign VP predicts its machine learning features will offer an alternative to A/B testing in e-mail marketing - November 15, 2019
- InsideView product exec says Data Integrity will address B2B firms’ cleansing and visualization needs - November 14, 2019