Using AI for customer analytics casts a new light on the question, what (the hell) is quality all about?
The quality culture has its roots in manufacturing. In a perfect manufacturing line, a perfect product is the reward for a perfect quality. Perfection, free from faults and defects, is the quality measure in an engineered world. There have been many attempts to use this understanding of quality also in customer analytics. Why not measure the quality of processes between supplier and customer and identify ‘defects’ (as an offset to the desired) to eliminate. The problem has been that what was ‘perfect’ for one customer was not perfect for the other (other than with machines). In addition, interaction in a moving eco-system is different to the production flow of a manufacturing line. Nothing is linear in customer analytics.
Why not use Artificial intelligence to connect all this non-linearity? Once you succeeded in building a data structure that supports you and is able to digest your data, you encounter a next obstacle. You can build models that are very precise, or models that are very general. In the sense of ‘quality’, we are happy to encounter a path to perfection, and therefore we go for precision. But only to be surprised to see that the ‘perfect’ model only is perfect for that case in that moment. When time elapses and conditions change, perfection is lost quickly. Moreover, the ‘perfect model’ is not perfect for all the other customers.
What happens, if we go with generalization instead of precision? In that case we do not reach the level of 100% precision, but only 85%. There is a remaining error rate of 15%. However, this error rate enables the model to learn, because learning requires an unbiased contrast of right and wrong.
Standing in the shoes of a neural network, trying to become a more and more powerful one, learning is what is essential. It does only work, if the world is not perfect. (We probably do not have to worry too much about this point.) AI works well in customer analytics, if the focus is not perfection, but adaptation. In that light, the ‘new quality’ is the quality of adaptations to changes in the ecosystem.
What we take away is that the understanding of quality – if you want to use AI extensively for customer experience analysis – must be a different one. It is not any more the zero-defects philosophy. In social networks, it is not perfection that counts, but adaptation. It is about something like a ‘lifelong’ learning process of companies. However, if that is the case, then errors and defects are a ‘must’ for all kinds of quality.