5 minute read

AI Should Be Helpful, Not Hurtful

Each year the buzz around AI and machine learning builds as the must-have technology brands need in order to excel today, remain competitive, and meet rising consumer expectations. And most companies are (literally) buying into the buzz: 78% of brands have said they’ve already or are planning to implement AI by 2020.

But the problem with AI is that if not approached or leveraged properly, it can end up hurting instead of helping your business goals, personalization efforts, and customer experiences. To get to the bottom of the most pressing questions and views on the topic, we sat down with our Co-founder and Chief Data Scientist Dean Abbott to share some perspective.

Dean clears the air on the most common misconception businesses make about AI and how to best approach these initiatives to make your AI initiatives helpful, not hurtful, in 2020. Take it away, Dean:

What’s the biggest thing to keep in mind when evaluating or implementing AI technology?

One of the most common questions I get asked each year is “How do you think AI will transform [insert industry] in [insert year]?” While I do believe AI will continue to push the boundaries of decision-making and marketing automation, AI itself is vague and misleading for many. Contrary to the hype, AI isn’t magic—if you sprinkle your data with a bit of AI fairy dust, it won’t start making you more money.

Leveraging these capabilities can be a highly impactful way to gain a deeper understanding of your customers and offer strategic personalization. But AI at its core is just math, and it won’t mean or do much of anything without the right people and data in place to power it. Successful AI initiatives cannot be achieved by technology or even by data scientists working with technology. Success comes from the collaboration between business stakeholders, data science, and IT (the three-legged stool I often mention, explained a bit more in this infographic) to ensure the right data is accessible to fuel your models and that the models are actually working to solve the problems facing your business.

I develop models at SmarterHQ to help our clients interpret customer behavior—their engagement and decay, likelihood to purchase, lifetime value—based on what happened yesterday, what’s happening now, and what happens in the days after their visit. This allows us to predict what they’ll do next and in return, exactly how and when to best interact with them in the future. But these models wouldn’t matter much without the ability to a) collect and process real-time behavioral and product data at scale, and b) collaborate directly with clients to tweak these models based on their specific needs and the ways in which their customers uniquely interact with their brand.

So if there’s one thing you keep in mind this year, it’s to understand that AI alone won’t transform your business. Ensure your teams and/or vendors are experienced and actively working together to achieve the same goals, and allow for iterations and input along the way. That’s where the real magic happens.

Top 3 priorities marketers should tackle to make their AI initiatives more effective in 2020?

1. Data Accessibility: Having access to the right data is key to making any sort of machine learning capabilities work at all, let alone be more effective. But there’s a reason why data collection, unification, and integration is a recurring priority and work in progress for most marketers: it’s hard to do across internal teams, data streams, and interaction channels—and those who say they’ve already prioritized this most likely cannot do it at the level of scale or accuracy they need to just yet.

The size and complexity of data and consumer expectations continue to rise each year, and so should the expectations you have of your teams and technology to evolve with it. Streamlining your data and making it immediately actionable should be a priority, always.

2. Customer Intent: Many marketing decisions are based on average shopper behavior, but when you lump everyone together, you often miss out on key patterns of behavior that deviate from the average. Or as Sam L. Savage wrote in his book, The Flaw of Averages, “plans based on average assumptions are wrong on average.” For example, one might find that shoppers with nine or more visits in the past week have a 10x increased likelihood to transact online in the week following—but the marketing message will be different depending on which nine products the shopper viewed, or whether the shopper transacts regularly or if these visits are their first experience ever on the site.

Make it a point to prioritize a more accurate and holistic understanding of customer intent in order to send the message or offer that makes the most sense to each individual based on their specific behavior, where they are in the customer journey, and engagement level.

3. Historical View: Tracking real-time behavior is important, but so is incorporating the customer’s past interactions and purchases. Make it a priority to focus on more than just what the customer is doing right now. Pay attention to what the customer also did last week, last month, and last year when trying to interpret what the action they’re taking now means.

How can brands combat the “creepy” factor associated with AI/marketing personalization?

Transparency. CCPA, GDPR, Apple’s Intelligent Tracking Prevention—they’re all a direct response to giving today’s consumers what they want: more transparency and control over their data and brand interactions. Your customers just want to have a choice.

If done the right way, personalization should never be creepy. However, the reason brands employ “creepy” tactics is because it generates revenue. But eventually those short-term wins create burnout and your long-term strategies and customer relationships suffer. Brands need to respect a consumer’s right to choose, and they also need to understand that every customer is different: their motivations and interests, what they consider personal vs. non-personal information, their communication preferences—which is exactly why we need data and personalization, smart models, and smart people working together to accurately cater to each of them.

As AI continues to evolve, it’s been easy to build models that make personalization more effective and less offputting, but we have to continue to build models and employ campaigns that are more relevant instead of blasting to the masses. If you’re not customizing to your customers, you’re underserving them. The whole point of our segmentation capabilities is to take this broad, on-average behavior and slice it down into pieces that can distinguish and then suggest content, recommendations, and cadence for that particular subgroup of people. And then you can better personalize and interact with these people without them feeling like you’re crossing into creepy territory.

Ultimately, consumers want brands to understand them better, but they also want brands to understand when to leave them alone. When Alexa reminds you to order dog food? Helpful. Alexa listening to a conversation about what to wear to a wedding and then you receiving ads based on that conversation? Creepy. Use common sense—and ensure the right tools/technology are in place to accurately target and deliver 1:1 experiences—and it won’t be creepy.

For more on how to be transparent and prioritize strategies that are respectful of what today’s consumers want, download our Privacy & Personalization Report.