Conversational AI won't help you if you don't wanna talk

While all posts written here have it implied that the opinion is mine and not my employer's, I would want to assert it more explicitly in this post.

In professional capacity I have been working on products to help businesses automate their contact center interactions via Conversational AI (voice bots specifically) for quite some time now. I understand that there is value in this. One could reasonably say that it's the reason why I am in the field. But, over two days last month, I encountered four extremely poor experiences—as a user—with customer contact centers involving some kind of Conversational AI. All these companies had decent, if not good, customer care experience before. And it was extremely disheartening since it made me question whether I am looking at the problem correctly.

Any customer care system would have a set of allowed solutions to problems and a way of discovering and executing them via conversing with an agent. I am using agent generally to also cover app interfaces that have much higher efficiency when the problems and solutions are standard. While considering Conversational AI deployments, customer care teams have the responsibility to ensure that, as a whole, the number of allowed solutions, discoverability, and execution efficiency either improves or stays the same¹. But most fail miserably at this.

For many clients, customer care is a second-class citizen and they don't really have the will to talk to the users. And this lack of will reflects in how they look at adoption of new systems. The four interactions I mentioned earlier in the post all had effectively blocked talking to human and were probably proud of 'AI adoption' while reducing their conversation system to something a 6-button-app could do since the edge cases were not allowed anymore.

Of course it's the fault of us vendors too. With human contact centers no one was deconstructing 'customer care' deeply enough since many things were hidden away in the norms of human interactions. While offering a replacement, the onus is on the vendors to provide a complete picture. But I guess it's hard to move in a principled manner in such bull markets. Most clients look at cost reductions and it works out well enough to get and sustain deals. Don't get me wrong here, while it's challenging I definitely know that a few true customer-first companies are not messing this up. I'm also hopeful for the future, as sometimes these 'low priority' problems are inadvertently solved through general product upgrades. But I will stand by my core premise that if you don't want to really talk to your users no AI will help you.