Contact center AI: Agents may be training their replacements
The goal for software developers training sophisticated agent-assist AI tools for contact center software platforms could be 'superbots' far smarter than today's chatbots.
Contact center AI technology innovations are coming quickly, as platform vendors improve machine learning and language tools. The spotlight is now on agent-assist tools that listen in to calls or follow agent prompts to crawl customer service records and suggest answers to customers' problems during a live conversation.
Moreover, augmented reality (AR) tools that were on display at the ICMI Contact Center Expo conference earlier this year also show promise for enabling customer self-service for walking through the setup and installation of hardware, such as cable modems. They can also add a visual element to live-agent interactions with customers and suggest fixes to technical problems, which verbal descriptions over phone or chat can't replicate.
But industry insiders are uncertain about the development trajectory of the genre of agent-assist tools.
One well-informed view of contact center AI suggests human call center agents are simply training their robotic replacements. The humans are curating massive data sets that will be used to train future generations of more sophisticated bots, similar to how Google users click on CAPTCHA images not only to solve today's web identity security issues, but also hone the cloud giant's AI image recognition engine.
These superbots could feature more advanced natural language processing intelligence that will conduct more complex conversations, elicit more detailed information and solve higher-level problems that today's bots can't fix.
Tomorrow's bots -- trained on data already being collected by human agents rating how helpful each successive suggested answer is -- may end up being the cost-cutting technology and layoff trigger some employees fear.
ICMI leader not seeing it
That hasn't happened so far, said Brad Cleveland, ICMI's co-founder and past president, in his keynote address at the May 13 to 16 contact center conference in Fort Lauderdale, Fla.
Artificial intelligence is called 'artificial intelligence' for a reason -- it's not a relationship.
Brad ClevelandCo-founder and past president, ICMI
Existing contact center AI tools can automate some low-level tasks Cleveland characterized as ones that "shouldn't have precipitated customer contact in the first place," often because of a product or service's insufficient development or documentation. It takes customer service engaging with customers to unearth those issues so they can be corrected. As far as bots replacing human contact center agents, for now, Cleveland said "we're not seeing it."
"Artificial intelligence is called 'artificial intelligence' for a reason -- it's not a relationship," Cleveland said. "Whenever that's necessary, contact centers are going to be important as ever."
One measure of contact center AI's progress is how bank customers rarely call to ask for their account balances as they did in the past, said Jafar Adibi, head of artificial intelligence and data science at Talkdesk, based in San Francisco. They just log into a website or ask a bot. Agents will have to be trained and empowered to solve more difficult customer problems as bots soak up the easiest work, he said.
Agent-assist tools and augmented reality are the latest examples of AI in call centers.
"Minority Report, that kind of stuff, we'll get there," Adibi said, referring to the 2002 science fiction action movie directed by Steven Spielberg, in which police catch criminals based on human predictions about the future.
Adibi said, in 10 to15 years, bots will perform many more tasks than they do today. It will take that long to get them up and running, because there are many "simple pains" in contact center AI technology that still need to be eliminated before the next generation of bots can realize more humanlike attributes.
Adibi added, though, that some customers will likely always prefer talking human-to-human for some customer service problems. But some members of younger generations, he noted, actually prefer interacting with bots.
Pointing the camera at the back of a device, such as Wi-Fi routers or cable modems, in one customer application, shows data about what cables to plug in and where. The technology can also guide agents to figure out what's gone wrong by connecting them to documentation and offering suggestions for solving problems.
The startup sells to both contact centers and field service operations, and it supports live chat, visual databases to more quickly identify devices and problems, screen sharing, reporting and analytics for service engagement.
According to Tom Tseki, vice president of sales for North America at TechSee, the idea behind the technology is to prevent "truck rolls," or service calls at residences and business, or when those visits happen, assisting agents with likely fixes to the problem precipitating the call.
"If I'm going to throw a chatbot in front of my customers, it only covers a few use cases," Tseki said, adding that AR tools may represent the next generation of customer service for some companies' goods and services. "It's only going to answer 10% of the volume, and now the other 90% are pissed off because they wasted a bunch of time with the chatbot and still have to go talk to somebody."
AR tools cut technician dispatches to customer homes and offices by 19% and product returns by 17%, according to a recent survey of 60 TechSee clients covering 220 contact centers and 30,000 agents. Customers also indicated they increased first-contact resolution -- or solving a customer's problem on their first call, a key measurement of success for many contact centers -- by 22%.
