The next generation of easy
The journey to developing an intelligent Easy Button began with investigating various technologies and platforms to better understand what kind of software was currently available and what it was capable of doing. Cognitive computing surfaced as the most exciting technological advancement.
“When we started looking into cognitive solutions, a senior product manager who is also very technical took it upon himself to demo the Watson platform,” recalls Bartley. “In a single afternoon, he put together a simple but powerful conversational platform. It was a wake-up call for us—that cognitive solutions are real and the tooling around them [is] powerful. That’s what got us really interested in Watson.”
To be thorough, the team examined other cognitive platforms from major market players and tried to project where those platforms might be in 1 – 3 years. The IBM Watson Developer Cloud platform, with its extensive set of services, tools and support, stood out. “We felt that IBM had, by far, the largest lead in terms of where cognitive was going and that the Watson team would be in the best position to help our business users,” says Bartley.
Today, the enhanced Easy Button is part of a larger Staples Easy System, an intelligent ordering ecosystem that allows Staples’ business customers to order supplies quickly and easily from a variety of devices using voice, text or email input. The system takes advantage of the natural language processing (NLP) and machine learning capabilities of the Watson platform as well as numerous Watson services. The technologies work in unison to understand and process customers’ spoken words, translate speech to text, extract the nature of their requests, and use voice feedback to provide order confirmations and product recommendations.
The Easy Button itself is outfitted with sensors and wireless networking technologies that communicate with the company’s ordering, commerce and customer data enterprise systems through an underlying integration platform. The IBM Watson Conversation service, a platform as a service (PaaS) technology, puts a conversational interface on the button so that customers can interact with it using natural language. The Watson Speech to Text and Text to Speech services work together to process customers’ utterances and enable utterances from the Easy Button device.
When a customer speaks into the Easy Button, Watson Conversation works to understand the customer’s intent and entities referenced. Currently, the system is trained on five intents, or skills: product ordering, product reordering, shipment tracking, checking on reward summaries and processing back-to-school lists from scanned images provided by customers. Entities are office products contained in the company’s vast catalog, such as pens, toner and paper, each of which has its own stock keeping unit (SKU).
After Watson Conversation recognizes the intent and entity of a request such as, “I want to reorder black pens,” it calls the Staples personalization engine. That engine uses the Watson Retrieve and Rank service and custom-built analytics to comb through the customer’s purchasing history and identify the specific SKU, or in this example, pen, the customer historically orders. If the system is highly confident it has identified the correct SKU, it uses voice feedback to confirm the purchase with the customer, and the transaction is complete. If the confidence score is medium, the system suggests a variety of pens based on past orders so that the customer can select the correct product.
To provide customers with anytime and everywhere service, Staples is also using its Easy Button software platform with Watson Conversation to support all of its chat experiences across its many channels, including the Staples website, the company’s chat and mobile app, and third-party messaging platforms such as Slack and Facebook Messenger.
“Watson Conversation helped us immensely and changed the trajectory of the project,” says Goodwin. “The visual tooling around it made everything significantly easier—easier to train Watson on our product catalog and intents, and easier to see where we were having issues. It’s fantastic.”
The Staples Easy System with Watson also learns over time. For instance, if it cannot confidently return a customer’s request—perhaps because it hasn’t been trained on the intent or the customer makes a request it can’t understand—it forwards the request to a live agent who can help the customer. In cases like this, Staples feeds the agent’s conversation back into the Watson system so that it can learn the dialog and successfully manage similar requests in the future. “We’ve set up an end-to-end learning system so that we can continue to improve all of our chat and voice channels around specific intent,” says Bartley.
Anywhere commerce and service
The ultimate goal of the Staples Easy System and the Easy Button is to remove friction from the ordering process and enable office managers to place orders wherever, whenever and however they want. Goodwin explains: “Calling customer service and waiting on hold, or logging on to a website, searching through rows and rows of products, checking products out…it can be a hassle. Ordering supplies should be easier. The Easy Button makes it simple and magical. You tell it that you need pens and it knows exactly what you mean.”
Bartley sees using cognitive capabilities across the Staples brand. “Anywhere we’re building or using software to help our customers or associates, we’re thinking about how cognitive services from Watson can be at the heart of that software product. We think it can support many different spaces.”
For instance, by integrating NLP and machine learning technologies with its existing big data and personalization platforms, Staples can gain a deeper understanding of customers and their preferences. This kind of thinking has Staples leading the B2B market in what it calls conversational commerce—the ability to deliver anytime, anywhere service and commerce capabilities using voice, text and email.
Staples is rolling out the system in phases, starting with launching it on its mobile chat channels. Currently, more than 10 percent of Staples’ Apple app users are interacting with the company using chat. With the introduction of Watson technology, Staples expects that number to increase. It also predicts higher order frequency, larger average order sizes and improved customer service scores.
In October 2016, as part of its alpha testing, the company will offer a pilot version of the system to a select group of customers and use the resulting feedback to inform product and software design and further train the system on use cases. Even in the project’s early phases, demand for the Easy Button is high. “We already have large enterprise customers reaching out to become Staples customers, mainly because of the Easy Button. They want it in the office and to be part of it,” says Goodwin.
Beta testing will begin in November and focus on augmenting the system based on customer experiences, releasing a companion app, and enhancing hardware and software with 3G wifi and 2-way voice capabilities. It will also publish the system to other chat channels across Staples. Over time, as customer requests evolve, Staples will add new skills to the system. “Eventually, we think there will be 5 – 10 dominant use cases, but we’ll let the customers decide which are most relevant to them,” says Bartley.
In the future, the company plans to augment its system by incorporating the IBM Watson Tone Analyzer service to better understand customers’ emotions, personality traits and language styles, and the IBM Watson Sentiment Analysis service, part of the IBM Watson AlchemyLanguage collection of services. The API will help internal customer service agents gain insight into customers’ attitudes and opinions.
“A year ago, we wouldn’t have been able to do this,” says Bartley. “Because of the work of IBM and others, software development is at an inflection point where cognitive services are now available to everyone and can power unique and differentiated experiences.”
Working with IBM and the Watson Developer Cloud hosted on the IBM Bluemix® platform has been—and continues to be—a positive experience for the team. “We’ve had some really good conversations with not only the Watson team, but also the commerce teams and leaders at IBM,” concludes Bartley. “The Watson team is very interested in our use case because we’re pushing some of the boundaries of how to connect directly with customers.”