I think about Conversational Commerce and its impact on Customer Experience for some time now. How should we develop AI powered conversational commerce technologies in a way it helps companies better understand the jobs customers are trying to get done. If done well AI is part of a great set of conversational commerce capabilities to humanize digital and facilitate seamless journeys across channels. You can follow the links if you want to read my thoughts on those topics in more detail.
And there’s more to think about. Here’s a couple of related thoughts.
Thought I: What if everything can be automated?
The AI (academic!) specialists think that AI will develop into a level that it can perform all human tasks in 50 to a 100 years. Many human tasks and jobs will likely be taken over much faster than that. And I firmly believe there will be more new jobs, but that’s not the scope of this post.
Until recently we treated the human as the default interaction mechanism. This is shifting towards seeing the human as the default point of interaction when technology (or process) fails. And that remains the situation for some time to come.
But, what if everything (including service recovery) can be automated? At what point in the customer’s journey, do we need to design in human to human interaction? What customer desired outcomes relate to the specific traits of human to human interaction and should therefor not be automated. Even though it can be? Do we know? Do we know how to know?
Three guidelines to decide when to choose human interaction over automation
It’s a discussion with a lot of emotions, assumptions and beliefs that make it hard to come to a conclusion. That’s why I think we need to structure the discussion. Here’s some “guidelines” to make a decision when – or when not – to design human interaction into your conversational commerce strategy:
- You should design in humans when your customer really needs a human. That’s likely a lot less often than your customer would say themselves. Nevertheless, you need to find out when (=in what context) your customer needs are better met when humans help to fulfill them (see though II for more on this).
- Then you should design in humans when doing so demonstrably and measurably creates more value for both the customer and your business, over not designing in the human. Maybe even if this does not help the customer to get the job done better?!!
- And, last but not least, you should also design in a human for when technology fails, assuming that will still take place, even when all human tasks can be automated.
I think these cover the spectrum of options. Which brings me to the second line of thought
Thought II: How to test for a situation that does not exist?
We are living in the age of idea-generating abundance as the cost of testing them has dropped significantly. Companies like to develop and innovate by experimenting these days. The A/B test has reached peak popularity because we can now, measurably, say what works best, among different possible (and available!) solutions. But it is impossible to A/B-test a fully automated service, as it just doesn’t exist. Full automated conversational commerce is not available and therefor cannot be tested against human-to-human interaction.
And there’s another side of the challenge. Testing solutions is great, but testing solutions for real problems customers face when trying to get their job done, is even better. Experiments will only produce insights to the extent of what you are experimenting. Nothing more, nothing less.
So, how do we find out when to design in human-to-human interactions in a fully AI-powered humanized digital world?
What we should do, imho, is use Job-to-be-done research to understand customer needs and opportunities for innovating around those (unmet) needs. Once you understand customer needs you need a service design research approach to deeply understand customer’s steps and desired outcomes when getting their job(-steps) done. Then you should hypothesise multiple solutions (with and without human-to-human interactions), prototype and test them with real customers.
Of course this does not scale like A/B testing in a live environment, but it is probably the best way available to get early insights into why, in what context, for whom and how to design in human to human interactions in a fully digital world. Human-to-human interactions that get the job(-step) done better than purely automated interactions, even though powered by human-like conversational AI.
Thought III: What new automated interactions will we design?
We humans are very much programmed – by life – to see technology dominantly as something to replace humans with, to increase productivity and to reduce error or undesired variety. It is what we have seen most, it is what we know that works, it is what we can build a business case with. As I discussed above we are entering an age in which we need to actively understand when to design in human to human interactions.
But, these AI powered conversational technologies also hold a promise of designing new kinds of automated interactions that no-one has done before. Interactions that create new kinds of value for the company and the customer. Here’s a couple I can think of:
Pro-active inbound service conversations
In the future we will likely see an increase of interactions that AI powered conversational technology can make easier and cheaper to scale. The warm welcome call for subscribers to services is one that’s high on the wish-list of many marketeers. Yet e-mails show poor opening-rates and cold-calling these customers is an expensive exercise. AI powered conversational technologies can start these automated conversations on inbound contacts (e.g. when a customer comes to the website) with an option to escalate to live when needed.
Real-time customer feedback & recovery
Collecting customer feedback has long been the domain of surveys, online feedback platforms and communities. The real challenge many voice of the customer programs face though is that this is all a-synchronous. It is feedback after the fact. Nice for purpose of analysis, not so nice from the customer’s perspective of actual recovery. With AI powered conversational technologies you cannot only collect feedback when the event happens, you can automatically start-up the service recovery process. This automated conversation uses and captures important context to understand and solve the issue at hand. On top of it, this enables an entire new stream of valuable customer experience feedback data, which will help you better understand what customers want and how you can fix it. Because it is one to recognize when something goes wrong, it is another to act on the signal and do something about it.
Predictive failure discovery and service “recovery”
Even better would be not to have the problem at all. The best service is no service! The predictive nature of machine learning might prevent failure from happing and solve the issue before it occurs. Maybe by pro-actively sending in a human to get it done. The customer may actually value that more than having it fixed automatically, without him being aware there was a problem in the first place. Because customers appreciate it when a company visibly makes an effort for them.
These are my current thoughts on conversational commerce and the decisions we need to make. More to come in the future, for sure.
Please share yours.