Conversational AI continues to mature for customer service
As more businesses use conversational AI bots as the first line of customer service for consumers, the experience can usually be helpful or maddening.
Ganesh Gopalan, CEO and co-founder of conversational AI provider Gnani.ai, said Corporate AI That he understands these reactions, but insists that in two to three years, conversational AI bots will be dramatically improved and inspire more confidence in consumers and business users.
Early conversational AI systems were built around keywords, which, when mentioned by a caller, then invoke an associated response, Gopalan said. The problem was that this process often lacked context and did not adapt to regional dialects or informal formulations. For users, this could be frustrating because when the bot didn’t understand them, they were transferred to a live agent and had to repeat everything they had already shared with the bot. Exasperating.
By the end of 2021, the success of conversational AI for consumers depends on what they ask it to do for them, Gopalan said.
“If you’re going to make an appointment, it works today,” he said. Even booking a test drive for a car and other tasks can be done using bots, he added. But problems can arise if the client’s request does not meet the expectations of the AI bot.
“If you’re looking for a test drive at a car dealership, you can tell it in a million ways,” Gopalan said. “You can say you want a test drive, but you can also say ‘I just want to try a Ford Mustang’.”
Unless the bot’s conversational AI is trained for such a variation, it won’t be able to understand what the caller is asking for, he said. “The way we currently code this stuff is that you come up with the initial set of methods and then feed it to the natural language processing algorithms. They would generate similar sounding phrases, but this might not cover all options for all regions and there might be some peculiarities with some use cases.
On the flip side, for many customer service calls today, conversational AI bots work well about 70% of the time, he said.
“I think all of the routine stuff can definitely be done today,” Gopalan said. “If someone calls an insurance company to find out their policies or status, it can all be done. But if a customer calls in to complain that their claim isn’t covered and they want someone to just listen on the other end of the phone, I don’t think it’s completely ready today.
So when will more responsive and understanding conversational AI become possible?
“In a few years, I think it should get better and better,” he said. “What happens is the system learns from mistakes, so maybe you start with a system that works 80% of the time. The problem is, when companies don’t fix the remaining 20 percent, the system gets confused. You need a learning system, and you need NLP or a business like ours to focus on specific use cases.
This kind of customization for each use case is essential to enable conversational AI that is precise and usable for a wide range of uses and industries, Gopalan said.
“Everything has to be personalized on some level,” he said. “Just take something out of the trade and try to plot two things together, that won’t work. You aren’t going to anticipate everything that is going to happen, and patterns that generate new sentences are not going to generate everything you need.
And while conversational AI continues to mature and improve, it can never fully replace human interactions with a live customer service agent, Gopalan said.
“People are going to call and complain, or they are going to have complicated issues to solve,” he said. “You will always have human beings doing this customer service job. “
Rob Enderle, senior analyst at Enderle Group, said Corporate AI that the technology has been around for a while, but companies, including Gnani.ai, are working to cut costs and make it available to more companies.
“The technology is relatively mature, but rather expensive in its mature form,” Enderle said. “IBM Watson was used in production for insurance sales, and it was so believable that some of the men who were called attempted to ask the virtual woman who they were talking to. [on a date]. The problem is no longer the performance of the equipment, but the level of effort it takes to train it.
This training has been very laborious and expensive, Enderle said, and the hope is that the next generation of neuromorphic computers will reduce the time and associated training costs.
“Conversational computing can be successfully implemented today on a sufficiently large budget, and the industry is working to reduce that cost to something much more reasonable,” he said. “Once they’ve done that, probably in the next three to five years – and implemented it as a cloud service, this technology should become a lot more mainstream.”
The conversational AI market “represents the next big step in the human-machine interface,” said Enderle. “When it comes of age, it promises to radically change the way we interact with computers, putting a lot more communication load on them. This capability makes conversational computing a critical step in creating the future of computing and therefore one of the most critical efforts currently in development.
Gnani.ai recently announced a partnership with global contact center and business process outsourcing provider Transcosmos to deliver its suite of conversational AI products and services through the Transcosmos call center network. The move is intended to help Gnani.ai expand its presence in North America, the companies said.
In July, the AI conversational intelligence platform Chorus.ai was acquired by ZoomInfo for $ 575 million. Chorus.ai uses machine learning to pull together new data insights from traditionally untapped data workflows within organizations so that they can help drive additional sales efforts.