One of my favorite subjects in business and technology is the notion of disruptive innovation. According to Clayton Christensen, who pioneered the term, a new technology or product becomes disruptive when it addresses the needs of customers at the lower-end spectrum of the market, usually ignored by existing players who are focusing on the high-end, high-margin customers. The disruptive product or technology is sold at a much lower price point, and offers a lower performance, at least at the beginning. Later on, as the technology improves, the disruptors are gradually moving up the chain, increasingly pushing incumbents into smaller, high-margin niches.

A notable example of disruptive innovation is the open-source database software. In its early days, it was only used by hobbyist developers or small companies who couldn’t afford a proper enterprise-grade database from Oracle or IBM, often sold for 6-7 digits figures to large corporates. As the technology improved, more and more of the larger companies began to adopt open-source databases, at a significantly lower cost than the enterprise-grade software. Now, anybody can spin off a new database server (on-premise or cloud) at no or very little cost.

So what does this have anything to do with AI and customer support? In the past, AI technology was only accessible to large companies with big budgets who could afford to hire expensive machine learning experts and had access to powerful compute capacity. That changed a lot in the last few years, mostly thanks to open-source software but also to cloud service providers like Amazon or Google. It is very easy now to use cloud services such as Google’s Dialogflow to create natural language models which could accurately and automatically categorize any piece of text at the fraction of a cent each.

Unfortunately in customer support, AI is often considered a high-end tech, available only to teams with large budgets. The costs, in the range of thousands of dollars per month, are prohibitive to smaller or medium-size companies.

The affordable AI exists though and is already widely available, as mentioned in the example above. In my view, this is an integration problem. That is why I decided to write a step-by-step tutorial on how to implement your own ticket deflection project at the fraction of the cost of using existing customer service AI vendors.

I hope it’s useful.

Sorin