Can machines deliver the ultimate customer experience?

car-smlMy immediate reaction on hearing CEOs talk about delighting their customers is to cringe. As true delight is so rare, this catchphrase largely comes across as inauthentic. But when companies get it right, the impacts are incredible.

I recently went back to a Sydney hotel I used to frequent when I ran my own business during the first dot-com era. To my surprise over 10 years later, it was the same cultured Italian concierge that greeted me. But my greatest surprise was that he remembered me; asked after my family and my business, introduced me to the new management and ensured my room was upgraded. Now that was delightful. So guess what, it’s now my Sydney hotel of choice.

But how does an organisation scale this type of experience to ensure that delight is the norm, not an aberration? In an offline world, you’re dependent on the quality of your customer service team and the training you give them. And just how many people can a concierge know that well?

In an online world, we can augment some of these limits through technologies such as CRM and cross-channel marketing platforms. But the reality is that we’re doing pretty poorly, judging by engagement rates in our emails, display ads and social channels.

The reality is that human beings have hard coded limitations on the number of relationships they can maintain. In the late 90s, anthropologist Robin Dunbar at University College of London, published a paper called “Co-Evolution Of Neocortex Size, Group Size And Language In Humans”  in which he proposes that the mean group size for humans is 147.8. This also matches census data on village and tribe sizes across many cultures.

The average number of friends held by a Facebook use has grown to almost three hundred as of August this year. The averages vary significantly by demographic, but for the purpose of this argument, it is tempting to conclude that the utility of Facebook has enabled us to break Dunbar’s limit. But I think we need to look below the surface to understand what’s really going on.

Here’s a simple test: go through your Facebook or LinkedIn connections, and ask yourself how many of these you could confidently recommended a book, movie or restaurant with a high degree of confidence of getting a ‘like’. I did this with my 500+ LinkedIn network, and came up with an answer of about 100.

This is hardly scientific, but more robust research is definitely supporting the view that if you consider true relationships only (connections with a high level of emotional intensity over time), Dunbar’s limit holds. The reason is relatively straight forward; most strong relationships are built and reinforced with nonverbal communication, which is why we still meet face to face in preference to videoconferencing.

Coming back to customer service, I’m certainly not advocating picking the fleas off your customers to build better relationships. But there is no doubt that CRM and marketing platforms have improved customer service, by augmenting the agent’s ability to retrieve customer history and simulating a knowing relationship. So this works in human-mediated environment like a call centre. But can this scale even further by removing the human altogether?

Earlier this year, IBM released a customer-service version of its supercomputer Watson, which became famous when it beat two champions from the American game show Jeopardy in 2011. The “Watson Engagement Advisor” will be able to respond to consumer questions and hold a basic conversation by keeping track of context and history. By crunching big data in near-real time, Watson promises to transform the way brands will engage with clients.

Based on the success of its field trials, several large brands have now committed to rolling out Watson, including ANZ Bank (Australia) and the Royal Bank of Canada. Whilst the first application will be at the customer service end, one can see the direct applicability to guiding purchase decisions and ultimately 1-1 marketing. Watson is currently text based, but as speech recognition and output technology develops, we’ll probably get a version of Siri that’s actually useful.

We are close to a technology tipping point. The ability to process big data in near-real time, coupled with cross-channel marketing platforms that are aware of the consumer context (you’re on our website, or in our store) and communicate using the appropriate channel on the right device at the right time based on the consumer’s need, will be the end of direct marketing as we know it.

When I finally receive an email that feels like it’s been written just for me, and guided through an online purchase journey by a virtual agent that I really can’t tell is human or not, we will have finally passed the Turing Test and have built a marketing brain that can actually think as well as my favourite concierge.