What is AI / Machine Learning in the Service Management Vertical

Back in 2001 we had integrated a natural language knowledgebase engine into our application when Google was still a garage start-up and Alta-Vista was all the rage (remember those days). This KB engine was a key architectural component to enable our clients to build their own collections of documents and allow natural language search for “self-service” (remember when that was a buzzword).

Anyway, after a few years of algorithmic tweaks, the inclusion of meta data and the real-time indexing of the request repository (ticket data) we realized that the KB engine got “smarter” and more accurate with its results. We then started to store more meta-data, specifically around categorization and Eureka!!!! We had no administration, “machine learning” , intelligent matching of requests from any inbound channel – email, phone, service portal etc..

The end result is that IncidentMonitor^TM learns based on your historical data to aid you and your user community to correctly categorize your inbound requests every time; significantly improving your metrics and statistics through data accuracy – and we’ve been doing it since 2006. Additionally, you can quickly identify trends, Major Incidents and/or Problems that are contextually related.

Now is that AI and Machine Learning or is it just a robust architectural framework? Some vendors would call this AI/Machine Learning but to me true AI would deem a Service Desk redundant. Just think, if we had true AI we would have an FLR of 100% and a Re-open Rate of 0% and operational costs of cents/hour. Is that what they are really offering or is it just a new wrapper on an old technology that they are finally catching up on.

I’m not sure if Chat Bots and Virtual Service Agents answering questions are truly going to provide any value-add when it’s much faster to simply use a known method of Natural Language Processing to search content directly.

Before the use of any "Bots" we need to have our content organized with meta data so it can be searched naturally - this is the first and most important step. Then we can think of engaging Chat Bots or Virtual Service Agents. It’s like anything Garbage in, Garbage Out. I’ve seen some predictions stating that the majority of AI Service Desk implementations will fail because of the lack of content for intelligent responses. A Chat Bot without NLP indexed data is like asking Barney from the Simpsons how to drive a car – answer “BBRRRRRPPPPP”..

So next time when someone touts AI and Machine Learning in the service management arena be skeptical and get the details – like all of the other technologies before it , it’s not a silver bullet but another tool in the tool bag.