For Chatbot Press 1by Dave Michels in Telecom
Chatbots are the new IVR. That is chatbots are a technology that will transition from promising to imprisoning.
I’m not against chatbots. They are a very clever evolution of the auto attendant. They have lots of great applications from helping hospitalized children inquire about their menu to selling Taylor Swift CDs in Asia. They will become very popular. So popular that, like IVRs, we will learn to hate them.
When IVRs were new they were revolutionary. Seriously! Back then a lot of calls were routine requests that could easily be automated. This was pre-web days, so the phone was the primary tool to getting answers to basic questions such as operating hours or directions.
The IVR was a result of several breakthroughs including touch-tone dialing that enabled end-to-end signaling, new inexpensive digital signal processors for audio, and microprocessor-based platforms. The IVR rapidly evolved. The spell by name directory routed calls and computer-telephony integrations (CTI) made it possible to do balance inquiries and other tasks.
The IVR ROI had a cost cutting driver (receptionist time), but there were also adopted to improve customer satisfaction. For the first time customers could get answers to their questions 7×24. Applications such as a dial-registration for university classes provided immediate confirmation on student schedules instead of week-long delays associated with bulk ScanTron processing. The IVR was the tool for customer enablement, satisfaction, and liberation.
Many people find that hard to fathom. IVRs today are generally reviled. We immediately press zero or say “representative” before the first option is even presented.
Enter the chatbot. It too is being pitched as a revolutionary solution that will drive customer enablement, satisfaction, and liberation. Chatbots can be created with web, app, or speech interfaces, and are possible because many customer inquiries are similar and predictable. This time the breakthrough technologies are messaging, web-design, and artificial intelligence.
The IVR was all about self-service. The chatbot is all about conversational UI. It’s a nice Star Trek-like concept, but still relies on self-service. Since the prompts are not simple digits, it requires some AI. There lies the IVR 2.0 problem: AI doesn’t work very well.
AI has come a long way, and it’s almost useful. My problem with AI systems is they assume too much. A good example is my Google Now phone tells me when it’s time to leave the airport. It’s fantastic, almost useful technology that factors in current and predicted traffic patterns, my location, and the fact there is a flight in my calendar.
While Google Now knows a lot, it doesn’t know how I plan to get to the airport, where I plan to park, or if I intend to check a bag. If it did know or attempted to know this inforimation it could be useful. In my case, I usually take the bus (something Google should actually know). It could/should suggest the time to leave for the park-and-ride and know which bus I need to catch. That would be useful.
The problem with AI is it never asks. It’s a badge of honor to assume correctly – though rarely does. We are being asked to believe that these chatbots will interpret our requests accurately. No chance considering that I have trouble interpreting the text messages from my kids and spouse – and we know each other and speak the same native language.
The other issue is that most things in our modern life are complicated. This is what caused IVRs to go south. It was simple to offer helpful responses like business hours, but simple requests no longer exist because they get resolved by websites. Today’s calls (and chats) are complex. IVRs were great for account balances, but not very good at troubleshooting or correcting complex situations. When was the last time you called an airline? It wasn’t to check a route, book a ticket, or check flight status. We call airlines when something is really messed up, and even then it can be a burden to get past the IVR. Do you really have high hopes for the United Chatbot?
The added complexity of a speech interface is fodder for another post.