Predict & Influence Muses - Series 005: Confusions Between AI and Analytics

While the popularity of this buzzword, analytics, or big data analytics, is more than 10 years old, and for the word AI, maybe 5-8 years old already, there are still a lot of confusion or mixing with these 2 concepts, or use these interchangeably. This is not correct. I have also seen even IT practitioners making the same mistake too.

It is never too late to clear this misconception. For a start, AI is mostly about prediction. Banks use AI to predict which transaction is fraudulent. ?Google uses AI (more specifically, knowledge graph-powered recommendation engine) to predict what are we trying to search or know. Grab and Uber use AI to predict the best match between the most suitable driver and passenger. The list can go on and on. AI, and its branches, which includes the much talked about machine learning, deep learning models, natural language processing (NLP), computer vision, speed recognition etc. are continue to get better and better in terms of ability to predict better (ie more accurately), through numerous innovations and breakthrough from the academia and tech giants (which are able to spend top dollars to hire distinguished AI scientists).

AI is also mostly linked to automation. The results of the AI prediction will be automated to be consumed or used by users. For example, we use the Google search user interface to perform the search, and the results of the AI predictions will then be displayed as search results, through the ranking of relevancy.

Analytics, using big data (can be loosely defined as using structured and unstructured, and internal and external data alike if we don’t go by the Gartner’s more accurate definition using the 4Vs, which are Volume, Velocity, Veracity and Variety) or otherwise, is different. Analytics, while it may not to be known at that time, has a much longer history. We human want to make fact-based decisions, and since school days we have been taught to use 5W1H, the Who, What, When, Where, Why, and How to understand a situation. Analytics is supposed to use data to help to answer such questions as: “Who are my top 10 customers?”, “When did the user logout from the system?”, “What products did the customer buy?”, “Where did the passenger go after this?” …. In this situation, Analytics would present the insights but still require a human (i.e. senior management, analysts etc.) to understand, interpret the insights, and eventually, make decisions.

The hardest questions to be answered are the Why and the How. ?It is no surprise that the answers for the Why and the How typically requires unstructured data, and some forms of prediction (read: AI) will be needed. For example, if we want to answer the question “Why the company’s sales has dropped?”, most likely the answers does not lies on the company’s structured data sources like ERP alone, we may need the sentiment data collected from the CRM system (i.e. complaints around product quality has been increased), external chatters (i.e. news, forums, public social media pages) that a competitor has launched a new product and that it is well received in the market. These insights can be formed through “Structured the Unstructured”, through firstly AI, like using NLP, and then presented in the form of dashboard (a form of analytics manifestation) so that the user of the dashboard can find out the possible reasons for the Why.

Using the above example, I have shared the possibility of AI and analytics works together in a use case, i.e. for text analytics using NLP (AI) and then present the finding through Analytics. With this post, hopefully you can distinguish better of AI and Analytics the next time someone asks you ??