Cutting-edge companies are turning to artificial intelligence and machine learning to transform how they interact with customers, strengthening those relationships, distinguishing themselves from competitors, and increasing revenues.
At the center of this transformation are technologies such as chatbots, recommendation engines, personalized communications, intelligent ad targeting, and image recognition. Gartner analyst John-David Lovelock predicts that AI-derived business value will more than triple to $3.9 trillion by 2022 — with improvements to the customer experience key to this growth.
With the promise to improve customer acquisition and retention, artificial intelligence is attracting the attention of business and IT leaders alike. According to Salesforce Research survey of 3,500 global marketing leaders, 51 percent of companies are already using AI, and more than a quarter plan to deploy it within the next two years.
Here is a look at how organizations are putting AI to use in improving and hyper-personalizing the customer experience and increasing their customer bases.
Chatbots
When it comes to customer experience, the poster child for AI use is the automated agent, virtual assistant, or chatbot. In practice, the technology remains rudimentary but companies are making good use of these technologies, often for very basic customer support, as long as humans stay in the loop.
At Pearson Education, more than 2 million people log into the company’s systems every day, with on average 15,000 people per day running into problems they need help with — a number that goes up substantially during back-to-school season.
Difficulties accessing specific educational materials, problems with their browsers, password resets — these are the problems Pearson decided to address with AI-powered chatbots this past April, says Steve Santana, the company’s CTO for customer platforms.
Targeting the back-to-school season, Pearson extended its relationship with Salesforce, by deploying its Einstein chat capability.
“We got into an early pilot,” Santana says. “And we were able to get it up and running and implemented in six weeks.”
The impact was bigger than he expected. More than 60 percent of the 63,000 access, browser, or password help request were solved by the chatbots. The other 40 percent were rolled over to human agents, who were able to resolve issues 20 percent faster because a chatbot had already captured much of the needed information, he says. Pearson will also use these calls to help improve chatbot performance.
“It was a huge time saver for our learners and educators, and for us,” Santana says. “It was fewer agents I had to pay during the key selling season.”
Pearson is now working to add more types of questions into the system in time for the next back-to-school period, in January. It is also considering using chatbots in the sales process, such as validating professors and checking on the status of their orders.
Santana recommends taking time to understand chatbot use cases and making sure chatbots are trained well.
“It’s important to analyze the calls that have been successfully handled by agents,” he says. “Find the best agents, and have your chatbots trained by them.”
According to a recent Constellation survey, 50 percent of companies are using AI in customer service projects, either in pilots or production. But some companies are waiting for chatbot technology to improve.
“I do think there’s a future in it,” says Jacob Wagner, IT director at Bluestem Brands, which has 13 different retail brands, including Fingerhut and Bedford Fair. “There’s some level of triage, hypothetically, that chatbots can do. And that level gets more robust as the technology grows. But people really like to talk to other people.”
One day, he says, chatbots will be good enough that customers won’t be able to tell the difference. But companies shouldn’t expect to replace their entire support operations with chatbots.
“Companies are getting value with the simple questions,” says Ray Wang, principal analyst and founder of Constellation Research. “But you can’t just replace your customer care center.”
In fact, as chatbots get better at answering the simple questions, people will need to get better at dealing with more complex problems, he says.
Recommendation engines
Bluestem’s Wagner may be holding off on chatbots, but his company has seen noteworthy success applying computer intelligence to another popular area: shopping recommendations.
It’s not enough to remember that a particular customer likes to shop for, say, musical instruments, Wagner says. “I’m not always going shopping for the same reasons,” he says. “Sometimes, I’m shopping to buy a present for my wife. Constantly showing me musical instruments is only interesting when I’m shopping for myself, not shopping for my wife.”
Understanding the customer’s context — what they are looking for in any particular shopping session — is essential. Here, drawing on the search experience of other users can help. If, for example, a customer who normally shops for musical instruments is in the sports section and searches for “driver,” artificial intelligence can help ascertain that the customer is looking for a golf driver, not a computer driver, which makes the shopping experience easier and faster for the customer.
Bluestem has been experimenting with recommendation AI for five years now, with three years in production. Wagner says pointing to concrete gains is challenging, because the market as a whole is changing. But he can measure the effectiveness of any particular AI change by running tests.
Bluestem’s conducts tests on 10 percent of its site visitors, with half of those receiving the current experience as a control group and the other half experiencing the new version. Once the conversion rates of the control group match those of the 90 percent who are not part of the test, the test has run long enough, and Wagner can see whether the improvement works.
“Generally, depending on the metrics, it’s small gains, a quarter of a percent conversion increase. But we do it all the time,” he says. “We have two or three tests live at any one time.”
As the system matures, the company explores additional factors that could influence a customer.
“Layers of complexity start to occur to you,” he says. “For example, people who are spending the weekend in Cancun have different shopping habits than when they’re shopping while at home in Maine. So where you are, that’s important.”
Wagner would also like to use AI to improve product pages so that product information is presented in the way each individual customer prefers to see it. For example, certain television attributes are more useful to some people when making a buying decision, and should be placed higher on the page.
Personalized communications
Electronics retailer Adorama has been using AI for recommendations for two years, but its use of machine learning in generating marketing emails has provided the highest degree of personalization for its customers — and has saved the company money in the process.
“Before, I needed a copy writer to create subject lines,” says Lev Peker, Adorama’s chief marketing officer. “Now, with ML, I can achieve the same thing. So it’s solved some staffing issues.”
Adorama is using third-party vendors for its personalized communications campaigns. “The reason we like going with third parties is because we don’t need a team of data scientists, and we don’t need a team of Python developers,” Peker says.
For example, the company will add pixels to its web pages that send data to the third-party service provider, provide access to Adorama’s product catalog, and test the new system to make sure it is effective, a process that takes 40 to 50 hours.
Peker cautions against jumping on the AI bandwagon, however.
“You have to do this in a very smart and measured way,” he says. “Think about what problem you’re trying to solve, and whether AI can solve the problem, then measure if AI is the most effective answer, and the cheapest answer.”
For example, when it comes to personalized emails, using AI-powered technology from Persado has improved the click-through rate, leading to revenue increases of 50 to 70 percent, says Peker.
Persado trains its system on a company’s prior communications. It then takes a new marketing message and adapts it to be in line with the company’s larger marketing narratives. Results are measured over time to see how the AI-driven campaigns affect engagement and unsubscribe rates.
Nearly 60 percent of marketers plan to use AI in their content marketing strategy this year, up from 43 percent last year, according to a recent BrightEdge report.
But humans remain in the loop, says Assaf Baciu, co-founder and SVP of products at Persado. First, they create the core marketing message that is then personalized for each customer by the AI. And humans get to review the final messages. “The brand itself can see that the messages make sense, and they can click on elements of the message and refine them,” Baciu says.
Ad targeting
AI-fueled advertising is another major focus area for companies seeking to transform their customers’ experiences.
Consumer products giant Procter & Gamble, for example, has a whole internal data science team working on areas where AI can be applied, and advertising is a key area.
“We have a programmatic platform that understands who the customer is, analyzes if they’re the right customer to buy, what the right brand to advertise to them is, and how much to pay for the ad — all this in under a second,” says Guy Peri, the company’s chief data and analytics officer. “It’s a mass precision model, an example of AI applied at scale in real time.”
It’s also a more cost-effective way of reaching customers compared to its previous broadcasting model.
“These smart audiences are proving to be a more efficient use of our media spend, and the ads are more relevant to our viewers,” he says. “It not only improves the experience for our customer, but makes us much more relevant and efficient in what we’re doing.”
P&G builds its own AI systems in-house based on a blend of open source tools and commercial products, Peri says, adding that the company has expanded its AI strategy over the past three years to include teams in the U.S., Europe, China, and Singapore.
“AI is sometimes a shiny object,” he says. “But we’ve been very clear that at the end of the day, it’s about solving business problems.”
For example, P&G has been very careful about the use of chatbots, Peri says.
“We don’t want to jeopardize the consumer experience with bots that aren’t functioning properly,” he says. “They can assist human beings to better serve customers, but we don’t want to leave it all to a bot that may or may not be accurate.”
Image and video recognition
Image processing is another area where AI is helping to transform customer experiences, and online home furnishings retailer Wayfair is one company leading the way.
Wayfair started out with AI several years ago to help make smarter advertising purchases on websites, social media, search engines, and even television. But soon computer vision became particularly interesting because Wayfair’s product lines — furniture and home accessories — are visually-driven.
“These are products that are hard to keyword search,” says John Kim, the company’s global head algorithms and analytics. “With a rug, or a couch, it’s difficult to figure out what keywords you should plug in to get the products you want — but you’ll know it when you see it.”
Customers aren’t going to look through the 10 million products that Wayfair offers, he says. “We have an enormous catalog, and it’s super important for us to show the best products for each and every single customer through personalization.”
AI is used not only to recommend similar-looking products to customers, but also allows people to take pictures of a real-life object — a coffee table, say — and get suggestions for similar-looking coffee tables that the company offers.
It’s also used to recommend products in a particular style, such as modern, contemporary or industrial, and to figure out how best to present a product to a customer, Kim says. “What is the optimal picture to show — the product with a white background, or in an at-home experience? Is it a side angle, or a straight angle?”
Next, AI will be used to help customers update their home designs.
“You can say, ‘I have this room, and I want to make it more modern,'” Kim says. “This is not a current capability, but one we’re working on. I believe this is very cutting edge. Nobody is doing this. Nobody can look at a room and say it’s 20 percent contemporary, 40 percent modern and 40 percent industrial. Nobody is doing it, and we’re in a unique position to do it.”
AI-powered image processing also feeds into another cutting-edge technology that Wayfar is experimenting with — virtual, augmented, and mixed reality.
“They aren’t going to hit the mainstream for some time, but we’re bullish on those technologies, and are investing to make sure we’re ahead,” says Kim.
In August, Wayfair launched a mixed-reality shopping destination for the Magic Leap augmented reality headset, followed by an interior design and room planning app for the Magic Leap, announced in October.
P&G is also putting AI to use in image and video recognition projects.
For example, Olay skin product customers can use an AI-powered skin advisor via the web or a mobile app to take a picture of their faces and have it analyzed to learn more about their skin types and what products would work best.
“We use deep learning technology to analyze your face,” says P&G’s Peri. “It uber-personalizes the consumer experience, simplifies it, and drives repeats as a result of loyalty to Olay.”
Offline, AI-powered image recognition has helped P&G improve product displays on store shelves. The company is now looking at using AI for video analysis.
“It’s really important for us to understand how products are used,” says Peri. “Consumers allow us to videotape their actions, their use of Swiffer mops, or their use of laundry products. Today, processing through those videos is tedious at best. It’s challenging for humans, much less [so] for machines.”
Peri says he expects AI to be able to interpret actions more accurately than humans. “I think the evolution of image and video processing is going to be the next frontier.”
Toward an AI fabric
While each of these approaches to improving customer experiences involves unique use of AI, the reality is that AI-powered customer experience is becoming increasingly interrelated, says Wayfair’s Kim.
“Your pricing system has an impact on your marketing, and your marketing system has an impact on your demand forecasting, and the demand forecasting could have an impact on your promotions — and I can go the other way, as well,” he says. “They’re all extremely correlated.”
Now, he says, companies have the opportunity to combine these systems and optimize them all at once for a particular business objective.
“The main thing for us is that data science and algorithmic approach is really embedded in the entire organization, not just in a few key functional areas,” he says. “If engineering is the backbone of the company, data science is the life blood pulsing throughout the company.”