The Four Stages of a Chatbot’s Business Intelligence Evolution
I see four stages in the progression of chatbot-like AIs interacting with business systems for the purpose of providing actionable business intelligence.
Stage 1) Single Numeric Response
Question :
Bob, can you give me the order number that corresponds to customer number XYZ123?
Method: simple SQL query on one table. No need for translating back to natural language
Availability: Now.
Stage 2) Multiple Numeric Response
Question:
Bob, can you tell me if any orders shipped late yesterday?
Method: SQL query with some simple summarization. Possibly more than one table involved. Output will be in the form of a list
Availability: 6 months
Stage 3) Predictive Numeric Response
Question: Bob, how many of the orders we received today will meet the 2 day shipping window?
Method: predictive analytics applied to data queried from multiple databases. Response will need to be translated back to natural language to convey the nuanced nature of the response
Availability: 6 months to 1 year
Stage 4) Predictive Question Generation
Question:
Bob, what should I be worried about today in the shipping and customer service departments?
Method:
A combination of various machine learning techniques including deep learning (neural networks), analysis of unstructured data, k means clustering, regressions, and random forest.
Availability: 2 to 5 years