Bots, done right, are the cutting-edge form of interactive communications that captivate and engage users. But what kind of potential do they have for sales, customer support and the bottom line? Tobias Goebel, Aspect Software’s Director of Emerging Technologies, weighed in at our recent VB Live event.
Recently, I had the opportunity to exchange thoughts and opinions on artificial intelligence, machine learning in general, and chatbots in particular with Ron Brachman from Cornell Tech and Akhil Aryan from Haptik. It was all part of a VB Live event hosted by VentureBeat.
Chatbots have been mainstream in the news and in technology discussions ever since Facebook’s F8 chatbot announcement back in April — and there are no indications that chatbot chatter is letting up anytime soon. We had a lot to talk about in our session — and not enough time to cover everything — so I thought I’d try to answer some of the questions here that we didn’t get a chance to address.
What use cases is machine learning best for? Where are we most advanced? Speech, images, text? Something else?
To answer this, I should explain that my main area of focus is customer service, and that can be a delicate field. A lot of consumer engagement with a company happens when something goes wrong — such as a recently-purchased broken product or an incorrect bill or invoice.
By nature, these situations can be highly emotional. And as a business, you want to be responsive to potentially problematic customer inquiries like these. So relying on a chatbot to resolve issues that require a human touch might not be the best idea.
This is especially true if you let your bot “learn” from interactions it sees (say, in user forums) with no or minimal supervision. Things can easily go wrong, as the disaster around Microsoft’s Twitter bot “Tay” showed.
On the other hand, with the right supervision and enough training data, machine learning as an A.I. technique can help build very responsive and accurate informational chatbots — for example those that are meant to help surface data from large text collections, such as manuals.
I’d say that machine learning as a technique has been shown to work best on image processing. The advancements that Google, Facebook, and innovative startups such as Moodstocks (just acquired by Google) are showing in that space are truly amazing. Part of the amazement however, comes from the fact that we now see software take on another cognitive task that we thought could only be managed by humans.
What can bots do for my bottom line?
In my opinion, a bot’s primary application lies in customer service since most companies unfortunately continue to rely on an ancient methodology to manage customer interaction. And this is to be expected as most consumers themselves are still “hard-wired” to pick up a phone and dial a number when they want to engage with a company.
Companies haven’t necessarily made it easy for consumers to transition to digital-first interaction. Consumers are forced to either download a mobile app, browse websites, or use voice, the “dumbest” channel the smartphone has to offer, to retrieve information or perform transactions.
This is truly unfortunate because when it comes to paying a bill, checking on an order status, or reviewing account transactions, nothing is easier than sending a simple message. And with 900 million users now on Facebook Messenger, 1 billion on WhatsApp, and hundreds of millions more on basic SMS, companies have a consumer-preferred new medium for engaging with customers.
With messaging, a simple question can be posed in a simple message such as “Where is my order?”
Contrast this to the conventional options of being forced to shepherding that question through a maze of web or mobile app menus, or with IVR systems over the phone. Now imagine how a consumer-adopted, digital and automated interaction for simple questions vs. agent interaction over the phone could impact customer service and its cost. When chatbots handle the most commonly-asked questions, agent labor is reduced or redeployed to manage more complex and time-consuming interactions. Simple and moderate issues are resolved faster, leading to greater customer satisfaction and long-term loyalty. Bots can help deflect calls from the contact center and your IVR, which further reduces speech recognition license and telephony cost.
How do you see the evolution of bots moving forward? What events will shape the next 6, 12, 24 months? What are your biggest concerns with respect to this wave of excitement?
Many companies are already launching bots for customer acquisition or customer service. We will see failures, and in parts, have already seen some. Bots are not trivial to build: you need people with experience in man-machine interface design. But to quote Amara’s Law: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”
Bots are here to stay, and will be a great new platform and make things easier for all of us. But bots that try to do too much or set unreasonable expectations will slow consumer confidence and acceptance of them. What might help us now is maybe to calm down a bit with the hype, and focus on building good bots that have value — then share our experiences, and show the world where the true value lies.
It’s all about expectation management.
Hear more from Goebel and others in this VB Live event and:
- Understand the AI outlook
- Recognize the key players (what they know that you should too)
- Anticipate changes in the AI landscape and plan accordingly
- Find ways to integrate AI into your business plan
- Ron Brachman, IEEE Fellow, Former Chief Scientist of Yahoo!
- Tobias Goebel, Director of Emerging Technologies, Aspect Software
- Akhil Aryan, VP of Product & Growth at Haptik
- Jon Cifuentes, Analyst, VentureBeat
Wendy Schuchart, Moderator, VentureBeat