Artificial intelligence is a wonderful thing. But perhaps contrary to popular opinion it's not a miracle cure or instant fix.
The dirty little secret of AI?
It takes time -- and sometimes a lot of time to get right. Perhaps even a year.
Rob Roy, the chief digital officer of Sprint, embarked on that journey over 12 months ago. Recently I interviewed him to learn how Sprint is using AI to personalize content and customer journeys for millions ... and significantly boosting sales.
John Koetsier: What were your initial goals?
Roy: "We wanted to deliver a digital differentiator for the enterprise using a data-first approach. That includes our website and apps ... and we run the full Adobe Experience Cloud stack to leverage AI across all of our properties.
That includes testing and personalization based on machine learning. We're leveraging the backend machine learning to understand segments of customers, what they're doing, and the likelihood they'll do something else. Then we serve up appropriate content and experiences."
John Koetsier: Can you give me an example?
Roy: "One example is product recommendations. Customers might be thinking: what device is right for me? And we may serve up something totally different from what they might expect.
Previously we would just guess, but what we're able to do now is leverage machine learning to pick a device that will work for them with a high degree of accuracy. That's helping us achieve a 14 percent lift in conversion rates, so now we're expanding it across other portfolio opportunities such as our marketing pages.
Now we're running multiple tests across those experiences to provide the best content based on history, lookalikes, and so on. Before there was a limited number of things we could think of, but now we can throw 100 different variations of content out there and let the machine decide which to show, and in what combination."
John Koetsier: How many segments are we talking about?
Roy: "Well, we're running at millions scale ... not tiny numbers of people. So most marketers try to work out their segments to the Nth degree ... maybe hundreds of segments.
But at even five to ten ... you're really just as good. Then you layer in the by-the-moment personalization, and we're able to personalize to a much wider scope while still keeping the number of segments manageable.
That's the beauty of AI ... you put a little bit of faith in the machine and let it do its work."
John Koetsier: Talk to me about how long it took you to put all of this together.
Roy: "The one thing about AI is that it's not a light switch. This is not the 40-yard dash, this is a marathon. We spent a good 12 months preparing.
It's like building a house. You don't just move in the day the foundation is laid.
The best thing that we did was spend that time preparing ... what it really did was enable us to move much quicker when we released it."
John Koetsier: How much quicker?
Roy: The foundation we laid spending that time getting ready has enabled us to get the flywheel moving. Before we used to do one to two tests each month. Now we do between 15 and 20 base experiments and another magnitude more subsequent tests.
John Koetsier: That's impressive. But taking the time to do it right requires significant corporate support, no?
Roy: "I have an amazing team. My faith was baked in their faith, and I was the face of the confidence to the leadership.
I had an honest conversation with our executive ... if you want it done this week, it won't be done right. The fortitude of our leaders to understand that was awesome."
John Koetsier: Thank you for your time!