Cognitive Robotics: Mind the machine

A recently developed subset of Robotic Process Automation (RPA), Cognitive Robotics (CR) is a much more advanced form of robotics. In  CR a computer can be “taught” to recognize patterns in data, and then to extract, classify and structure the data.

Using machine learning and user driven training, CR is able to harness the value of unstructured data (e.g. data that is disorganized or unreadable in PDF’s) and improve straight through processing (STP) rates by enabling such data to be fed into traditional automation processes.

This is of tremendous value when current data statistics indicate that approximately 80% of the data held by organisations is currently unstructured and consequently all but useless unless alternative ultra-efficient methods are used to analyse it.

How it works?

Traditional RPA involves automating repetitive tasks by programming a computer to perform specific tasks if certain conditions are fulfilled. It is mechanistic and requires human intervention if it gets “stuck.” i.e. it can never progress past what has been initially anticipated.

CR by comparison, taps into cognitive capabilities such as perception, attention, anticipation, planning, memory, learning, and reasoning, and using the principles behind them to enable computers to “learn” self-sufficiently how to respond to different types of problems or situations without being explicitly programmed. This independent learning drives an ever-increasing STP rate – Joy to any operational lead!

Example: PDF invoices in Portuguese sitting in emails are read and classified, with exceptions being sent to user for direction/decisioning. A CSV file of invoices is produced for action by RPA or people.

Why we love it

CR is able to harvest the value of unstructured data for BI and operational purposes – it starts to enable unstructured data to be used as a strategic asset.

It greatly expands the scope of traditional RPA by helping ingest new data sources, and improves on conventional Optical Character Recognition (OCR) through the innate use of learnings.

STP rates are typically at best 30% on OCR but improve to above 60% on CR enabled RPA – it is another reason why we rate the IQ Bot developed by Automation Anywhere so highly.

Unstructured data is typically structured manually, requiring effort that is typically time consuming and produces error ridden output. Using CR eliminates the need for this tedious and frustrating work. In other words, using the mind of the machine frees up humans to use their own minds differently!