Today’s intelligent systems can learn from customer data to discover and provide insights that drive better experiences, heighten employee engagement, and inspire companies to innovate
Like it or not, humans and robots are increasingly being forced to coexist in the same environment, and there’s little anyone can do to prevent it.
This is becoming especially apparent in the CRM space, where customer service, sales, and marketing professionals are all starting to feel the impact of technological advancements in their respective fields. More software vendors are now offering software that incorporates artificial intelligence (AI), machine learning, deep learning, cognitive computing, chatbots, intelligent assistants, augmented reality, and a slew of other innovations, designed to perform the otherwise time-consuming tasks that needlessly tie up humans. The technologies also promise to process CRM data, along with other structured and unstructured data, more quickly to learn from behavioral patterns, discover solutions, and surface recommendations and insights to help customer-facing professionals do their jobs more efficiently and improve interactions with customers.
Though AI has been around for decades, the applications are coming out in droves. In April, an IDC forecast predicted that this year, worldwide spending on cognitive and artificial intelligence systems would increase by 59.3 percent to reach $12.5 billion. Of that amount, 9.8 percent is likely to encompass automated customer service agent tools. This suggests that these systems have really come into their own and can now be used by organizations of all sizes and experience levels.
There is a lot of buzz in the air, but the hype is not unfounded. To keep up in today’s competitive business landscape, companies that automate areas that can be simplified for the sake of a better customer experience will lead the way. And experts agree that all companies should at least be figuring out where to invest to get the ball rolling.
SO WHAT’S SO SPECIAL?
Michele Goetz, principal analyst at Forrester Research, says it’s important to understand what makes a solution “cognitive” or “intelligent,” and what doesn’t.
Chatbots, for example, have typically worked in a programmatic, or deterministic, fashion, meaning they can identify certain keywords or phrases that will indicate to which type of agent a case should be routed within the call center. A financial institution’s chatbot might pick up the term “loan” or “account balance” and then link that customer to the agent most qualified to help.
But “today’s intelligent chatbots—you can think of them as cognitive agents—are very different,” Goetz says. These systems, she points out, go steps beyond, as they can take voice-of-the-customer information, call center notes and recordings, facts, email info, and other components of a conversation before moving the case into recognition analysis. At this point, the assistant can understand the types of questions being asked, why they are being asked, when those interactions are taking place, and the results that they produce.
When a customer asks about his account balances, an intelligent assistant can reason that he might be interested in seeing his last five transactions. Or it might surmise that the customer could be concerned about potential fraud. Registering this, it can surface information to the agent who will take the call, or to the customer directly. “It’s much more evolutionary. It’s aware. It’s adaptive, and that’s where the intelligence comes from,” Goetz says.
This kind of functionality can be applied in many ways to benefit customers. A financial institution can use the technology to help customers with portfolio reallocation. In such scenarios, the intelligent assistant can guide the financial adviser as he works. It can observe the conversations, acting as a search mechanism that helps the adviser find the right information across different systems and resources, so he can make more intelligent recommendations faster. Instead of an agent having to write down the customer’s information during a call and get back to him in a week with a proposal, the interaction can now happen in near real time.
In many cases, all of this can happen behind the scenes, unseen by the customer. “The customer doesn’t necessarily know that this is happening,” Goetz points out. In many cases, too, the agent can train the machine, increasing the likelihood that the recommendations it makes are better and smarter over time.
Vince Jeffs, director of strategy and product marketing at Pegasystems, stresses the importance of the instant feedback loop in teaching intelligent assistants. There are plenty of cases where machines can’t quite solve these problems yet, he says, noting that some cases still have to be escalated to humans. “But they can be assisted by these machines, and the agents can guide the machine on whether or not its recommendations were good or not.”
WHAT CAN COMPANIES EXPECT?
While the buzz might signal something that is still quite a ways into the future, a reality in which many customer interactions are completed by machines is not that far off. Most companies—or at least those that don’t have astronomical budgets—shouldn’t expect a very high level of sophistication just yet. But there are many tools they can implement to save their employees time on routine tasks.
Customer support is a common area of investment, experts agree. A number of technologies can help “reduce the strain” on call centers as more people contact them for help and advice.
“If you can have an automated agent that is able to answer 50 percent to 60 percent of the questions in a timely way, provide good responses to customers, and doesn’t make them upset, that’s a win-win for everybody,” says Dave Schubmehl, a research director at IDC covering AI and cognitive systems.
Schubmehl points to Autodesk’s use of IBM Watson’s Conversation tool to develop a digital concierge as an example of a company that got it right. The agent, referred to as “Otto,” can handle 60 percent of web-based customer service inquiries. Autodesk improved support ticket resolution time by 99 percent and significantly upped customer satisfaction.
According to Schubmehl, to make it work, the company had to organize its knowledge base. The system was trained to handle common customer and partner issues, including resetting passwords or rebooting a program after it failed. More difficult cases that weren’t covered by the knowledge base could get handed off to a live agent.
“It is important we provide our customers with consistent quality paired with the shortest response and resolution time,” said Gregg Spratto, vice president of operations at Autodesk, in a statement. “Our collaboration with IBM Watson allows us to expand the Otto concierge service and deliver prompt, effective, and authentic engagement to our customers.”
Goetz mentions a similar case involving a company that sells insurance through employers and sees its heaviest traffic during open enrollment periods. In the past the firm had to train temporary staff members to handle those kinds of calls and answer questions about insurance types and policies. “The quality of customer service would significantly decrease, and it was inconsistent in terms of how customers were supported,” Goetz says. Furthermore, since they were contracted employees, the temporary agents were not as invested in their jobs as full-time employees. With IBM Watson, the company was able to move to first- and second-tier support levels when volumes peaked, pushing the more sophisticated cases to live agents.
Another major benefit of automation through intelligence is in reducing the “grunt work” and freeing people up to do more interesting work that requires more complex skill sets and critical thinking, according to Schubmehl. After all, it can get depressing just helping one customer after another recover passwords or log in to their accounts all day.
This is in line with research from an Aberdeen Group study commissioned by Inbenta, which found that companies incorporating cognitive technologies to support customer service interactions have seen an 81 percent improvement in employee engagement rates.
“There’s been a lot of conversation about virtual agents and artificial intelligence eliminating jobs, but what we actually saw is like any other tool: It’s going to be valuable in improving [agents’] performance and their interest and satisfaction while working,” says John Forrester, Inbenta’s chief marketing officer. “Instead of dealing with repetitive, boring questions that customers might have, they’d be dealing with something that’s more in-depth and engaging.”
While customer service jobs tend to have higher turnover rates than other fields, technology can help businesses with retention, as employees are likely to be more enthusiastic and eager to learn new skills and tackle greater challenges.
WHAT ELSE CAN IT DO?
Behind the scenes, cognitive automation can be useful as well. Robotic process automation (RPA) can learn routine tasks that service agents repeat throughout the day.
“Robotic process automation has been around for about 15 or 20 years, but it’s primarily been rules-based, where somebody programs the rules for the agent,” Schubmehl says. “Now we’re getting to the point where you can have AI sitting there, watching the person doing the work, and then essentially developing the script automatically that the robot should act according to.”
That’s a relatively new area, but a hot one. Technology vendors like Pegasystems, WorkFusion, NICE, Automation Anywhere, and BluePrism are all adding RPA capabilities to their tool sets. “I think we’ll see that happen more frequently in 2017 and probably 2018,” Schubmehl predicts.
And while the call center might be the most obvious area and the one that most companies will want to look at first, sales professionals can also benefit from tools that simplify otherwise tedious processes, thus freeing them to tend to other parts of their jobs. Schubmehl points to Conversica as one vendor leveraging AI to simulate human voices and carry out conversations in place of sales reps, keeping leads warm so that when they turn hot again, they can be routed to a live salesperson.
In marketing, AI technologies can be used to learn customer preferences and identify the offers, images, and deals that are most likely to appeal to customers and to identify whether they should be sent via email or other channels and devices.
WHERE DO I START?
Goetz recommends getting started by testing and creating proofs of concepts. She says that companies in retail, consumer packaged goods, or manufacturing have a number of consumable applications from which to choose. A pilot can cost $50,000 to $100,000; implementation could range from $500,000 to $1 million, she says, but the proof of concept has already been given in many cases, leading to return on investment as high as 20 percent or 25 percent.
“Many organizations right now are doing piloting and testing and really working out to see the feasibility, and that entry point tends to be very low,” Goetz says.
Steve Laughlin, vice president and general manager of IBM’s global consumer industry division, points out that his company’s Watson APIs can be accessed via the BlueMix app development platform to help create and test programs before they are actually deployed.
Among the uses he’s seen already, some companies are turning the APIs into programs to help them understand personality insights from written texts on social media or to analyze images that customers have already seen to determine the next set of images they will likely want to see.
Staples, for one, recently used IBM Watson APIs to create a real-life Easy Button based on its plastic marketing icon. After doing design thinking with B2B customers of all sizes, the office supply superstore chain understood that there was a demand for an office assistant’s assistant to help stock up on supplies. The button responds to voice commands. “Right now, you can go into the supply room and say, ‘Hey, Easy, we need more copier paper, and by the way, it looks like we’re short on blue pens,’” Laughlin says.
“A crazy thing happened, and people started asking Easy for things that Staples doesn’t do, and it’s created a whole pipeline of new ideas for them to work with,” Laughlin adds. “So it’s become not only an innovation for them to engage with customers, but it’s become a source of ideas from customers for potentially new innovations.”
While IBM’s Watson is likely the most prominent technology in this space, other solutions, such as CognitiveScale’s Augmented Intelligence, Digital Reasoning’s Synthesis, and Narrative Science’s Quill, are emerging as competitive options.
Schubmehl suggests several other considerations when starting out. The first is getting the right stakeholders in place and then getting them all together to identify the business processes or functions that AI will replace. With that step taken, a company can set up a shortlist of service providers and discuss the potential outcomes for the application.
“One of the things that you’ve got to remember is that if you’re going to do an AI system, it’s going to be based on data, so you need to make sure that you have access to all of the data that’s necessary to drive the system,” Schubmehl adds. “A lot of organizations don’t have any real strategy about how they keep track of their knowledge, so really getting a handle on that first is something they should think about.”
Then companies need to engage subject matter experts in the design and implementation process, Schubmehl says. If you’re designing for the call center, you need to have the call center people involved in the design and the development. “If you don’t, you’re essentially asking for trouble,” because the design might not align with their needs, he cautions.
Goetz agrees that the design element is key, noting that adopting artificial intelligence calls for a desire to provide a human experience. Companies should ask, she says, “how do I reproduce, or almost replicate, what a human experience is going to be?”
“There are plenty of instances where we, as humans, still want human interaction,” Pegasystems’ Jeffs agrees.
In fact, designing an AI system, Goetz says, should mimic the process companies use to design a customer experience program.
“Sometimes they speak, so the tone, the way that they speak—is it very formal, is it conversational, and at what age groups are they speaking—those sorts of things really go into creating the best success in adopting artificial intelligence for customer engagement and customer experience,” Goetz says. “You can’t underestimate the design component that goes into that, and that’s how businesses have to approach this technology. You cannot think this is just another scoring engine or analytic engine that runs under the hood of your CRM or website or advertising platform.”