Recently, we have seen artificial intelligence triumph over humans in Jeopardy and chess. And there is a growing presence of virtual assistants like Alexa, Cortana, and Siri that populate our computers, phones, and homes. It's only a matter of time before A.I.-powered assistants play a significant role in the workplace, experts say.
In fact, the global intelligent virtual assistant market is forecast to be worth $5.1 billion by 2022, up from - $600 million in 2014, according to Transparency Market Research. What are the potential benefits and challenges of giving smart virtual assistants a home in the enterprise?
The relationship between machines and people is becoming increasingly collaborative as machines gain the capabilities to perform more tasks and routine activities currently done by people. Consumers can ask Alexa, Amazon's voice-activated app, for the weather forecast, sports scores, or even to turn down the lights (if it's connected to a wireless lighting system).
"We're at a stage that I'd describe as 'front-end speech engagement,'" says Steven Thurlow, VP and global practice leader at Verint Systems, an analytics company that includes customer service solutions. "Machines now have the ability to recognize a natural language question and provide an answer from a knowledge base."
Even without speech recognition, virtual assistants can provide personalized services for customers by accessing reams of data. For instance, Hodges-Mace, an employee benefits technology provider, offers an interactive virtual assistant that accesses employee-specific data and eligibility rules to guide an employee through a benefit decision process. The virtual assistant can also personalize its messaging based on an employer's strategy.
"With the workforce becoming more diverse and remote, we are seeing employers ask for help in non-traditional communication channels and best practices in maximizing their touchpoints with employees," notes Kevin Andrews, CTO for Hodges-Mace. "How employers deploy these messages may vary, but one thing is apparent, we are being asked to leverage big data to formulate and deliver more direct messaging versus broad general communication campaigns."
Next Up: A Virtual Assistant Network
Beyond answering questions in a linear process, the next step for artificial intelligence is to access a wide network of resources to provide cognitive support. For instance, In the film Her, Joaquin Phoenix plays a character named Theodore Wombly who purchases an A.I.-powered operating system called Samantha (voiced by Scarlett Johansson). Wombly's perception of Samantha changes from secretary to advisor to girlfriend and he is startled when he eventually realizes that Samantha is part of an expansive network of A.I. systems that all provide similar personalized interactions.
While it seems unlikely that people will fall in love with a virtual assistant (at least in the near future), it's not difficult to imagine companies building a network of A.I. systems that communicate with one another. In fact, companies are already moving in this direction.
Last week, IBM and Cisco said they will work together to build apps that integrate Watson (IBM's cognitive system) and other IBM services with Cisco apps, such as the collaboration platform Spark and conferencing service WebEx. The new solutions will incorporate structured and unstructured data from on-premise sources and the cloud, explained IBM in a statement. The goal is to build applications that can analyze workflow patterns and provide insights through the course of daily activities and interactions.
For example, a financial advisor might meet with an investor over Cisco video and a Watson service will surface real-time data and insights related to their meeting while both users store files in IBM Connections and share them through Cisco's WebEx.
Additionally, Slack, the fast-growing workplace collaboration and messaging app, is reportedly connecting third-party business apps as a gateway to an artificial intelligence system that can answer routine questions and proactively seek out information in the workplace. In addition to tracking projects and answering questions like "what's my password?" Slack will eventually present employees with information it thinks they want to know.
"Slack will know the people you trust and the topics you tend to care about, and over time it will figure out how to better route information to you," Noah Weiss, Slack's head of search, learning, and intelligence, tells Re/Code. "It becomes a robot that's working behind the scenes on your behalf to find things you should know about but might otherwise never see."
Virtual assistants are only as valuable as the data they have access to. And as more devices and systems become connected through the Internet of Things, a virtual assistant can potentially connect those data points to provide highly personalized services, observes Jack Nichols, senior technical program manager at Interactive Intelligence, a business communications cloud services provider.
"Given the productivity benefits, it makes sense that all employees will eventually have a personal virtual assistant to interact with and one that can interact with other virtual assistants," Nichols says. For example, a virtual assistant that has access to your email account could curate them, read them to you, and mark action items during your daily commute. It might also remind you that an employee's birthday is coming up and connect with the employee's virtual assistant to gather meal preferences, calendar availability, and find a restaurant on OpenTable that meets those preferences. It might even monitor your vital signs and recommend taking a walk after several meetings while mapping out an optimal route. Finally, it could bring your connected vehicle to the front of the office at the end of the day and let your spouse's virtual assistant know your ETA.
Siri, Google Now, Cortana, and other virtual assistants already offer many of these features. The next step is to streamline and standardize those services to match each person's needs. "Given how fast AI is developing, I really believe the potential of virtual assistants is largely only limited by our imagination," Nichols adds.
The Data Conundrum
Before virtual assistants can deliver truly frictionless experiences, "there's going to be some battles about who owns the relationship," says Verint's Thurlow. "If Siri is talking to your banking app to perform a transaction, is it a bank transaction or a Siri/Apple transaction? And if the data is transferred from the bank's backend system to its virtual assistant out to Siri, who's protecting that data as it goes through? And if something goes wrong, who goes to court?" These are some of the questions that will have to be answered before virtual assistants can be commonly used, Thurlow notes.
"Data is the new gold," agrees Guy Levy-Yurista, head of product at Sisense, a business intelligence software provider. There's a need to connect as many data points as possible, but because it's so valuable and sensitive, there has to be a multilayered security system in place." Just as there are restrictions to the data that employees can access, companies will need to establish data governance and security measures for virtual assistants.
Companies will also need to establish a set of constraints around the type of data that virtual assistants are exposed to and how it's processed. A case in point is Tay, an artificial intelligence chatbot that Microsoft introduced on Twitter in March. The more you chat with Tay, the smarter it gets, learning to engage people through "casual and playful conversation," Microsoft said in a statement. Unfortunately, many of the conversations were hardly playful. People tweeted a stream of offensive comments at the bot and within less than 24 hours, Tay was repeating those sentiments back to users, illustrating the adage: garbage in is garbage out.
"A poor quality of data can significantly impact the quality of the experience, particularly with completely unsupervised systems," notes Mark Hanson, senior director and worldwide head of the Cognitive Innovation Group at Nuance, which has built an enterprise virtual assistant called Nina. There are several practices companies should keep in mind to create an optimal virtual assistant, he explains.
The first is defining "the domain of the types of things the virtual assistant will be able to address" and establishing a "canonical flow of an experience" so that customers know what type of experience to expect from the assistant. For semi-autonomous learning systems where the pathways are less defined, it is important to have "governance tooling, which allows you to see what a virtual assistant has learned and audit this learning potentially before these new capabilities are available to end users," he adds.
Gathering high quality data is also critical. Aggregating data from knowledge bases, content management systems, and websites are good starting points, but "one of the learnings we have found in the enterprise is that the answers to many of the questions customers are asking in the contact center are not actually neatly contained within a paragraph on a website or knowledgebase," Hanson says.
Some of the most valuable sources of information are high-performing and experienced agents who understand best how to interact with customers, he adds. Although virtual assistants have yet to match a human's ability to show empathy or intuition, agents can help companies design more engaging virtual assistants. For example, if the customer expresses dissatisfaction with a service, the assistant should be trained not to respond with an upsell script.
It's unquestionable that artificial intelligence and virtual assistants offer a lot of potential for improving our lives at home and at work. And while there are still many obstacles, a future where collaborating with machines feels natural could be here sooner than we think.