As leaders in business or IT, automation has always been understood as an approach to manage cost and increase operational efficiency. However, recent advancement and improved approach of automation has introduced another decision factor about timing and extent of investment of resources into this area. Keeping the Buzzword aside, while discussing with several senior business leaders about the real interest level into automation and AI, it was clear that there is increased level of awareness and interest in pursuing the offered benefit. However, there were many who wanted to wait a bit more to understand the possibilities of Return on Investment. There were many, who knew a little but there was lack of clarity about what to expect from the automation or AI for the money spent. I am trying to decipher some of the mysteries to enable others about what to expect and when.

I. Intelligent Automation: Low Hanging Fruit for Back Office and Repetitive tasks

IA which is also called Robotic Process Automation (RPA), is the first step of automation with very limited intelligence. It mimics user actions to reduce human intervention or workload. It should be seen merely as virtual worker, who has extremely limit¬ed intelligence and will operate based on provided business rules. Most of the mundane tasks, which doesn’t need great deal of expertise, can be accomplished by these IAs or RPAs. Increased productivity at lower cost is the expected outcome for the approach of IA. Capital market, banks, and insurance firms may consider leveraging this approach. However, other industries may also consider. Functionally trained robots will become the new virtual workers, which executes rule based processes. The approach primarily contains technologies to combine how a user interfaces with system and associated rule based workflows. IA needs minimal investment and can be added into the existing IT portfolio.

II. Cognitive Computing and Autonomics: For Transformational Process Change

Cognitive computing is an approach, which are several steps further to IA.

Apparently, the new machine intelligence software will make mistakes, just like we do, and we’ll need to be thoughtful about when to trust it and when not to

It includes self-learning systems, which uses data mining, pattern recognition and processing in business language instead of any programming language. In Autonomics, the aspect of self-remediation is also included apart from self-learning. These are the domains, where minimal human intervention is anticipated. A virtual support agent could be one such example, where the virtual agent is continuously learning and self-correcting based on high level policies and rules.

III. True Artificial Intelligence:

True Artificial intelligence offers a combination of IA and autonomics and beyond based on newer algorithm to define Intelligence as emergence-view. Ability to think based on previous patterns per data mining and self-correct as per the business rules provides an ability to go beyond the obvious due to availability of additional supporting data, advance algorithm, and supporting infrastructure. We can divide this into two distinct components; one is interface to collect data (examples could be various chatbots) and the smart agent, which interprets the data in the background.

I am sure about the obvious question of why now and what is so new about it, considering artificialintelligence is rather an old concept from 1956. Following three factors come to mind for its increased relevance now:

I. Decreasing cost of computing power: Compared to the past, cost of computing had decreased significantly, which makes data mining, and enhanced calculations much cheaper.

II. Emergence of Big Data: Big data as cogitative value addition to the intelligence was not available earlier. With increased availability of surrounding and relevant data makes the job of artificial intelligence simpler.

III. Advanced Algorithms: The newer algorithm to support artificial intelligence stems from the concept of emergence view of intelligence, which helped define a new approach for Artificial Neural Network (ANN). The interconnected neural network can be trained using different datasets using deep learning aspect of ANN. Some of the tech companies such as Apple, Microsoft, Google, and Facebook are using this for some time now.

A new paradigm

Machine intelligence is different from traditional software. Unlike traditional hard-coded software, machine intelligence gives only probabilistic outputs. We want to ask machine intelligence to make subjective decisions based on imperfect information. Apparently, the new machine intelligence soft-ware will make mistakes, just like we do, and we’ll need to be thoughtful about when to trust it and when not to. The idea of this new machine trust is daunting and makes machine intelligence harder to adopt than traditional software.

The question for us is to find the balance between going too broad and wasting time and effort in a bid to “learn about AI”, which runs the risk of betting on a solution and then trying to find a problem to solve, or approaching AI in a narrow fashion and missing the point.