Evolution of Salesforce Einstein AI

Introduction:

There is a lot of talk around Enterprise AI i.e. using AI at the enterprise level particularly in areas such as Customer Service and Marketing. Salesforce (Einstein.ai) is researching and enhancing the capabilities in 3 different fields: Vision, Language, and Voice.

Vision

Vision can be used for:

• Visual Description
• Visual Question Answering
• Visual Object Detection
  
  

Einstein Vision Rest APIs used:

• Einstein Image Classification (Categorize Image at scale)
• Einstein Object Detection (Extract and Contextualize objects in images)
  
  

Some applicable Business Use Cases:

• Social Media – Out of the box Solution for Image-based brand recognition
• Sales & Service – Visual Identification of Products and Parts to streamline the sales and service processes.
• Insurance – Automated Insurance claims by identification of damaged parts and severity
• Retail – Product Identification for field sales teams to streamline in-store audits.
  
  

Language:

NLP consists of both Natural Language Understanding (NLU) and Natural Language Generation. Salesforce is advancing in all the top Natural Language Processing (NLP) tasks:

• Natural Language Inference
• Sentiment Analysis
• Zero-Shot Relation Extraction
• Question Answering
• Machine Translation
• Common-Sense pronoun Resolution
• Semantic Parsing
• Goal-Oriented Dialogue
• Summarization
• Semantic Role Labelling
  
  

Unlocking Language Breakthroughs with Multi-Task NLP

Multi-task learning is a subfield of machine learning in which multiple learning tasks are solved at the same time. Some use cases:

• Model Optimization (Continuous improvement with each answered question)
• Zero-Shot Learning (Ability for machines to adapt to new domains and learn new tasks without specific training)
  General Language Understanding (Adaptable to slight variations in questions and tasks and can generalize to unseen classes).
  
  

Einstein Language Rest APIs used:

• Einstein Sentiment (Classify text as negative, positive or neutral)
• Einstein Intent (Classify underlying intent in a body of text)
  
  

Some applicable Business Use Cases:

• Seamless Support – (In-Service industry) using Salesforce Einstein Chatbots
• Actionable Notes – Taking notes using Einstein Voice Assistant leveraging NLP services
• Automatic Triage – Automatic Triaging of cases, when a case comes in, understanding the intent and routing it to the right department
• Survey Analysis – Analysing surveys and the sentiment in the surveys
  
  

Vision:

• Einstein Voice Assistant is the product to watch for in this field
• Currently voice tech. integration is available in services/applications and eventually, it will be available as APIs in similar ways like Vision and Language.
  
  

Resources available today:

Let us look at the resources available today to work with Einstein AI.

• Most of the API can be UI-driven; such packages are available on AppExchange for download. This can be used as a point and click instead of code.
• There is a feedback API as well, to ensure the model is improving over time.
• Free Sign up (till 2000 calls per month) & Documentation: eisntein.ai/products
• Model Builder: sfdc.co/ModelBuilder
• Resources for getting started:
• Beginner: sfdc.co/VizLangOverview
• Intermediate: sfdc.co/ObjectDetection
• Learn more about Salesforce Research at “Einstein.ai”