More than a century ago, the Russian mathematician, Andrey Andreyevich Markov laid the foundations for much of the machine learning used in today’s driver-less cars and chatbot technology. He studied 20,000 letters in Evgeny Onegin, a poem by the grandfather of Russian classical literature, Aleksandr Pushkin, to predict the probability of vowels and consonants occurring in the text.
It’s a pity he had not been around a 100 years earlier to advise Pushkin on the probability of serious harm if you challenge your love rival to a duel!
To understand if Markov’s Models are relevant to the success of your MSP practice, try answering the following question?
How many times has your customer service desk received a support ticket for a problem solved in the past?
The chances are, if you have been successful for any length of time, the answer is many, many times.
The Problem
As the service provider grows it can face the following problems:
- Repetitive tasks can lead to boredom and lack of employee engagement
- Simple tasks repeated over and over again can be costly to service
- It can be time consuming and expensive to capture and recycle knowledge
The Promise of Artificial Intelligence (AI)
The market opportunity for service providers is expected to grow but increasingly there will be greater competition for profitable customers.
AI can help stave off commoditisation and margin erosion with the following benefits:
- Done right, AI can improve the customer service experience
- Reduce the time spent fixing problems
- Allow expensive resources to be re-allocated to higher level processes
- Improve profit margins
- Give you a competitive advantage
What do you need to get started?
First of all you need a lot of data to train your machine learning engine. The second thing you need is an easy-to-use interface to your new features and services. This requirement has driven an explosion in the development of chatbot technology and programming languages such as Python.
There are two types of chatbot. Bots that risk trying to parse anything you type at them, and bots that limit your input to a few safe buttons or keywords. In the former, Natural Language Processing (NLP) is not looking for keywords in your text, like a search engine. Instead, it uses machine-learned pattern recognition to match what you say to an “intent” which has been “classified”, which means the bot has been trained to look for certain things.
Acquiring the training data and the time to program your chatbot may not be possible for every MSP but not to worry, the Predatar team has been busy working on a solution, for the benefit of our partners.
The Predatar Service Chatbot
It’s long been known that men don’t like asking for directions when lost. This is probably also true for IT experts when faced with a support issue. That’s why we have built up an extensive knowledge base so that he (or she) can ask the chatbot for help without fear of loss of face.
For every knowledge base article we have had to think “what question is this article helping to answer?”. We wrote multiple questions for each of the 2000+ articles. All the questions, answers and topics created the data set which was fed into our NLP engine to build a chatbot that can understand questions, rather than just keywords, and reply with the relevant knowledge base article(s).
NLP means that if the user does not input the exact question, the bot will still return the right answer.
Once launched, we will retrain the chatbot by studying the questions users are asking on a daily basis and storing this information in the database.
Finally
If you have been researching AI to grow your business, we would love to hear from you. Drop us a note at info@predatar.com or why not make an appointment to meet the team at IBM Think 2018 in Las Vegas from 19th – 22nd March.
