Conversocial CEO Joshua March explained how customer service AI can perform mundane tasks to reduce the human cognitive load and allow employees to work more efficiently.
Dan Patterson asked Joshua March, CEO and founder of Conversocial, how AI and social media can improve customer service and reduce cost.
Patterson: Artificial intelligence and machine learning are changing a number of industries, not to mention social media. Of course social media, but social media along with customer service for TechRepublic.
Josh, how is AI infiltrating social media to make customer service more efficient and less expensive?
March: Sure, and hey Dan, great to be on. A couple of years ago I was sitting at F8, which is Facebook's developer conference, when Mark Zuckerberg announced the launch of the bot platform. This was the kind of first bot platform for messenger. Everyone went really crazy over it, I'm not sure if you can remember all of the hype that was happening at the time, but Zuckerberg himself was kind of heralding a future without phone calls where every app was going to shift over to being a bot within a few months, where it was the end of human customer service.
The reality was, you know, much, nowhere near what the hype was and everyone realized pretty quickly that if you build a very basic, kind of rule based chatbot, it's not that effective for customer service. People tend to not like it, you can easily get frustrated and kind of upset people.
The whole area just kind of lost a lot of the luster. Now, since then we've actually seen some really big developments happen in machine learning and AI. A lot of businesses have been, like really starting to figure out what does work and what doesn't work when it comes to implementing AI and bots and machine learning into messaging.
We've also seen along side it this huge rise of private messaging and messaging apps for business. You know two years ago when they first announced that, messaging was still pretty small. Over the last couple of years, now messaging has really just taken over the world, in terms of how people communicate with each other, within businesses and from businesses to consumers. We have all of these things starting to come together where we're now starting to see how you can really implement machine learning, AI and bots, combined with human agents within messaging to really transform customer service.
We're starting to see a lot of really big progress happening in that space. Even today with kind of simple implementations, if our clients are able to save 20, 30% of all the inbound messages, it can be handled automatically. I think over the next few years we're really going to see a pretty dramatic shift in this area.
Patterson: Josh I'm glad you drew that distinction between the hype and the reality. Of course the realities of machine learning are, the potential is incredible but it's a long way from here to there. I wonder if you could help us understand what some of the challenges that businesses experience from 2015 until now. What changed? What has made that big jump forward, to make conversational AI much more efficient?
SEE: IT leader's guide to deep learning (Tech Pro Research)
March: Sure. So, I think the big mistake that people made initially was trying to build completely stand-alone chatbots. So they tried to create a bot that could hold the entirety of the conversation between, with the customer and the business. The fact is, even today with the most advanced AI, the most advanced machine-learning technologies, no bot is ever going to be able to handle the full complexity of any kind of customer service situation that could come up, especially for a big business.
So people kind of tried to build these bots and quickly realized that, while they could handle some certain very specific tasks or simple bits of a conversation, if you tried to leave them there too long, eventually it would frustrate the customer and cause a problem.
So it's really been, a big thing that's been a change of approach. Where, and our approach in this is to go you know what, the great thing about messaging, which is different than other customer service channels, is that it's asynchronous. It's kind of like texting a friend.
So you know, I'll send you a message as a business and if you respond in five minutes or 10 minutes that's pretty real time. I might get that message and maybe I'll respond in an hour. It's not like a live chat that's happening on the website or a phone call where you're trying to handle everything there and then. What that means is that when a message comes in from the consumer to a business, a system, an automatic system can go hey, we respond to this automatically, this is a really simple question that's come in, or is the beginning of the conversation and do I just need to ask a few clarifying questions to find out what kind of problem people have and collect information?
Then when the bot needs to hand over to a human, it can just do that seamlessly behind the scenes. If that means it takes another five minutes or ten minutes for a human agent to respond? It doesn't really cause any kind of an issue for the consumer on the other side of it.
That means that by combining automation with human agents you can create this really seamless experience, where you're having a bot that says it doesn't understand. You're never creating a situation which frustrates the customer, but instead every little bit of automation you add just speeds up the response times, increases the efficiency.
So through that model, you can create a better and better experience without, while also kind of lowering the cost of delivering service without ever frustrating the customer. I think that's the really key thing that started to happen.
Patterson: Yeah I love the idea that it was a change of approach, not necessarily a change in technology that caused iteration that lead to innovation and eventually optimization. I wonder if you could leave us with a forecast? Maybe looking in the next say 18-36 months, when it comes to the advancements of machine learning, natural language processing and AI.
Where is that in terms of business tech? Not just the trends but the realities of implementation and use of these new tools?
March: Sure. So I think over the past few years the really big change that's happened when it comes to the deep learning and machine learning, hasn't necessarily been that it's kind of suddenly become much more advanced, it's more that it's become much cheaper and faster to be able to use these techniques on really large data steps.
What we see today is that there are a lot of vendors, you know, like ConverSocial, and like other people in the market who are sitting on these really big databases of hundreds of millions, sometimes of conversations that have been tagged, that have Sensum added and all this kind of thing. Which are really incredibly rich data sets for learning from.
Over the next 18 months what we're going to see isn't necessarily a huge change in the kind of front end of like bots talking to a customer, but it's more going to be behind the scenes. Figure out how to automate a lot of things that were previously done by a human agent behind that. Automatically understand the context of a conversation, you know tagging it, routing it automatically, collecting information, putting it into a CRM system.
A lot of the kind of drudgery work, you know simple transactional work that human agents have to be doing that add to their day and delay how fast they can get back to a customer. I think we're going to see real machine learning, deep learning, added in that behind the scenes.
What that's going to do is just really increase the efficiency of customer care, it's going to make it cheaper and faster for businesses to deliver a better experience over digital channels. That's where the big shift's going to be.
- Special report: How to implement AI and machine learning (free PDF) (TechRepublic)
- Google AI can pick out a single speaker in a crowd: Expect to see it in tons of products (ZDNet)
- Machine learning: A cheat sheet (TechRepublic)
- Cloud AutoML: How Google aims to simplify the grunt work behind AI and machine learning models (ZDNet)
- Google uses AI, deep learning to predict cardiovascular risk from retina scans (TechRepublic)