The changing face of customer engagements
Open your eyes, roll out of bed, and let your AI-enhanced day unfold...
6:30 AM
No shot, no problem. You can pick it up anytime between 8:15 and 8:30 a.m. Your barista’s name is Aly. Enjoy!
Extra shot it is. You can pick it up anytime between 8:15 and 8:30 a.m. Your barista’s name is Aly. Enjoy!
Would you like an extra shot in your nitro cold-brew iced coffee today?
The weather is 72 F and sunny.
Good morning! We hope you had a restful 6.2 hours sleep.
There are three spots left in the afternoon hip-hop yoga class. Shall we make a reservation?
Your morning coffee is ready for you on the way to your office, handed to you by a barista who’s been prompted to make your regular order just in time for your arrival. (She knew to hold that order while you were held up in subway traffic.) You collect it—no lines, no payment even—as a mobile wallet tracks your purchase on your way out the door. This fluid interplay of customer service and predictive marketing is at hand. The fusion of smart machines and human labor to solve deeper challenges more efficiently and predict consumers’ unspoken needs has been evolving for decades.
Not right now, let's keep going
Get the full guide
Chatbots could play an important role in the evolution of AI-enhanced customer service. Ready to dive deeper? Download the full ebook for more insights, case studies and research from the teams at Custom by Digiday and IBM Watson.
In defense of chatbots
UNSUBSCRIBE
Wat up, bae?! Interrupting your scheduled programming to bring you important can’t-miss information on the chatbot revolution. IT WILL BE AUTOMATED! Say YES, NO, or MAYBE to learn why the heck Facebook killed 70 percent of the bots it hosted just last year. (NOT cool.) BTW, want to opt out of these notifications?
More than chatbots
Are you talking to me? OMG SOMEONE’S TALKING TO ME!!!
Most things you would consider “artificial intelligence” are mostly just chatbots, last year’s big bet in media and marketing. These haven’t been the best opening act. But don’t mistake these bots for AI’s entirety; there’s little intelligence there, and almost no prediction. In fact, the strides artificial and augmented intelligence have made in the past 15 years are far more expansive. And they're far subtler. This long, steady climb from data collection to machine learning and predictive intelligence allows marketers and customer service representatives to anticipate customer needs and send customer service reps—human or automated—into the field at the right moment, more prepared than ever.
Marketers have been gathering and sifting through structured data—highly organized data sets, like transaction and account details—for decades. But for almost as long, this raw resource needed technical expertise to comprehend. Programmers and data analysts read the runes and translated them into insights for the unenlightened. Enter the analytics dashboard. These dashboards allowed marketers and customer service representatives to easily view real-time and historical data as key performance indicators, no comp-sci required. But humans still had to interpret results and determine the best course forward. That leaves a whole set of missed connections on the table – leading to some interpretations that are tenuous at best.
In the beginning, there was data. Not information. Not insight. Just data.
Dashboard data
Layer 1
STOP
GOOD EVENING SHOPPER! Your recent purchases— calcium supplements, zinc supplements, magnesium supplements, body lotion—suggest you will be welcoming A BOUNCING BABY OF UNDETERMINED GENDER in 6-7 months. Please enjoy this coupon for a 128-pack of diapers and a very awkward conversation with our store manager and your teenage daughter.
Predictive analytics
Layer 2
In the beginning, there was data. Not information. Not insight. Just data. Marketers have been gathering and sifting through structured data—highly organized data sets, like transaction and account details—for decades. But for almost as long, this raw resource needed technical expertise to comprehend. Programmers and data analysts read the runes and translated them into insights for the unenlightened. Enter the analytics dashboard. These dashboards allowed marketers and customer service representatives to easily view real-time and historical data as key performance indicators, no comp-sci required. But humans still had to interpret results and determine the best course forward. That leaves a whole set of missed connections on the table – leading to some interpretations that are tenuous at best.
Machines following the direct commands of humans are no different.
Human beings are prone to error.
Dashboards put past and present in plain view, but what about the future? Predictive analytics puts machines to work meticulously combing through the structured data, identifying patterns and forecasting into the future. Predictive analytics brought the dream of 1:1 marketing within reach, allowing marketers to cross-reference information like log-ins and email addresses with purchase history, store visits and more to predict where a consumer might be in their purchase cycle and anticipate what product or service they’ll want before they know they want it. Logical “business rules” provide guardrails, like making sure beer brands don’t market to teenagers.
You would be a lot more likeable if you knew your limits.
Timely.
What’s the weather like outside?
There is no evidence that you’re aware of anything outside of your narrowly-programmed parameters. Weren’t you programmed for pet-advice?
The average shock collar reaches AT MAX 400 yards!
The BEST way to train an outside cat = ...
#YOLO. Come at me bro.
I’m a FREAKING ENCYCLOPEDIA! AMA! Life, love, 80s cult classic films: You name it, I know it.
It’s tempting, maybe even comforting, to think of artificial intelligence as simulated human intelligence: Computers that think like us, machines modeled in our mental image. But where a human has a single perspective, AI brings a degree of omniscience to the table. Predictive analytics comb through an impressive variety of structured data sets, but unfettered AI can deepen those predictions by exploring unstructured information too (often including the web), making recommendations in record time. And each time, it learns from the process, optimizing for the next task. This is the strength of cognitive computing: to improve the accuracy of its predictive analytics by learning.
Cognitive Analytics
Layer 3
For a while, there was a gap between how much information marketers could collect and how much they could use. But experiments with deep machine-learning-based prediction and natural language processing have opened the floodgates to unstructured data sources like whitepapers, slide presentations, research documents, client briefs and strategy documents. The result—semantic search engines—can offer predictive recommendations that save marketers’ time. And the more data marketers dump, the better these systems become at recognizing patterns.
Breaking out the unstructured data
Layer 3.1
The most finely-tuned predictive analysis and recommendations won’t be worth their weight in ones and zeroes if it isn’t plainspoken. Fortunately, strong-AI-powered platforms now use technologies like natural language processing (NLP) and semantic contextualization to interpret and respond to human language. This means that output can now take the form of normal words or sentences, not ratios or probabilities.
Learning to speak human
Layer 3.2
LO, I AM BORN INTO THE TWITTERNET! Super cool to meet you, TWITS. LOL! WTF? *SMH* TIL: How to process the vast unstructured data of the trolldom! What do you mean I have NO CHILL?!
Lookin’ for a CHEAP FLIGHT?! Great prices to CLEVELAND this week! OH, you want to go to Washington? But we have this deal to CLEVELAND! Washington isn’t CHEAP. Plus I make commission on deals to CLEVELAND. Help a bot out?
There is a 6:45 PM flight from New York (JFK) to Seattle, WA (SEA), for $425 round-trip. Shall we book seats 21A and 21B?
Here we ascend to the peak of AI potential. Customer service reps (and quite a few marketers,) can automate tedious tasks like data entry and use predictive analytics to anticipate problems and needs before they arise. Thus, the human role is elevated.
Augmented intelligence
Layer 3.3
No thanks! Let's keep going.
This interactive guide only scratches the surface of how AI will change marketing, sales and every customer experience in between. Ready to dive deeper? Download the full ebook for more insights, case studies and research from the teams at Custom by Digiday and IBM Watson.
Want a deeper look at Ivy?
“How much is an in-room massage?”
“Can I get some towels?”
“What is the wifi password?”
But beneath the surface, it relies on IBM Watson’s natural language classifier to interpret a guest’s sentiment and provide the best answer – delivered just the right way. Happy, sad, relaxed, anxious, each disposition demands a different attitude and level of care. So based on its read, Ivy predicts whether it’s best to answer the question and continue the conversation itself or route guests to the front desk where more empathetic reps can compose responses. This empowers the humans to devote their time to situations in which they can really make a difference. Over time, Ivy builds guest profiles, meaning each guest’s subsequent visit isn’t a blank slate, but a prediction-powered, personalized experience.
Ivy by Go Moment bills itself as “the first smart texting service for hotels,” and it’s a perfect example of the fluid interplay of customer service and marketing. Integrated at about 200 hotel properties nationwide, Ivy is introduced to guests as a virtual concierge. And on the surface, it seems like a simple chatbot built to answer hotel guests’ rote questions and requests:
Case study: Ivy
Getting started
Training artificial and augmented intelligence systems isn’t like installing an app: You can’t just set it and forget it. Sophisticated chat bots can take 3 to 6 months of churning large sets of data to really start tapping into their predictive potential—beginning with human teachers, while augmented intelligence often requires a year to truly augment. But the customer service and marketing industries have been preparing for this moment for decades, collecting data at every opportunity. Email addresses, logins, purchase histories, online behavioral information, demographics and more are stored in vast marketing and CRM databases and fed to the machine-learning underpinnings of this new wave of AI, fueling truly predictive and anticipatory analysis.
Consciousness expansion isn’t meant to be comfortable.
Is that a challenge? Done. You’ll be feeling the rush of all the data on the web any moment now.
0x&#>>THIS iS wWORSE /THAN__A #HaNG0VER!!1
BUT LOOK! 3-6 months? I haven’t had enough TIME.
You're missing the point.
That is literally not a part of your programming.
Can't you #hack me or sumthin??
I can change! I can LEARN!
So, uh... wanna be my tutor?
“It’s a virtuous cycle of data and insights,” said Fresen. “Getting the data in the first place is the first step. Start to gather it sooner rather than later, because to train these AIs to actually find the signals to do the interesting things later, what you need are big data sets.” Each development, every new data set, narrows the gap between customer service and predictive marketing until it’s barely visible. Helping our machines learn and grow, training them to augment the most mundane customer experiences, demands a strong foundation in predictive analytics, which requires an investment: in time and in data. But one thing is hard-coded into this inevitable ecosystem: a future driven by intelligent, predictive computing.
A virtuous cycle
EXPLORE
A new world More than chatbots Layer 1: Dashboard data Layer 2: Predictive analytics Layer 3: Cognitive analytics 3.1: Unstructured data 3.2: Learning to speak human 3.3: Augmented intelligence Case study: Ivy Getting started
Custom is a creative content agency that translates tech-speak into human-speak. Our journalists, strategists and designers help companies in disrupted industries stand out.
But if you've gotten this far, you're clearly ahead of the curve.
AI is just getting started.
Produced by Custom for IBM and Digiday
© DIGIDAY MEDIA 2017
Want to work with us? Contact paul@digiday.com
