What’s the Deal With AI Connectors

Posted: 2017-07-25

Author: Admin

Artificial Intelligence (AI) is changing our lives, and these days feels more omnipresent than ever before. From Siri to autonomous cars, AI is progressing at a rapid pace. Moving beyond its sci-fi-movie robot image, AI has become embedded in our daily lives, as we use Google search algorithms, book a quick ride with Uber, and manage our smart homes.

On the business side, AI has immense potential to catapult enterprises to a whole new level, where machines will augment human intelligence for complex decision making. I believe it won’t be long before we start leveraging AI engines on a daily basis to conduct routine business. This flavor of AI for the enterprise can be referred to as an ‘intelligent systems’ approach. The future of enterprise decision-making systems is here, with embedded AI being the core transformational force.

Having said that, quite a few companies these days boast about their AI capabilities, though all they have is an AI connector to devices like Alexa and Google Home, with the intention to equip their antiquated application suites with a natural language.

To the uninitiated, this may cause some initial excitement, but leads to profound confusion about the use and scope of AI in the enterprise, and the role of the so-called AI connectors. To those well-versed in the domain of AI and enterprise software, the superficial addition of AI connectors feels like putting lipstick on a pig. This, however, does not stop enterprise software vendors from beating their chests about developing AI connectors to outdated Business Intelligence platforms, attempting to give those platforms a fresh breath of air. To me, it seems like a cruel joke on customers, as they are being set up for huge disappointments.

We must ask then, given the enterprise context, what is the ‘real value’ of an AI connector, and what does it stand for?

Let’s use the analogy of walking into a dealership to buy a car. The dealer enthusiastically describes the great features of the car, but then hands over all the parts to you to figure out and assemble yourself.

If, in our analogy, we gamely attempt to assemble the car, at least we might expect an AI connector to be analogous to the engine of the car; but sadly, these AI connectors may not even compare to a smart windshield wiper. This is exactly what what you get buying AI connectors for an enterprise: lots of headache, minimal value.

Now, let’s imagine an enterprise, and assume that we have connected a Business Intelligence System with one of these AI connectors. What is the value you would expect from such a connection? To my mind, it will be the same old demonstration of KPIs and standard metrics, without any context or reasoning. Would you really call this AI? At best, I would call it ‘rehabilitated Business Intelligence.’

So what is the real definition of Artificial Intelligence for business?

An Artificial Intelligence (AI) system should exhibit human-like intelligence and behavior, self-learning ability, and automation of business processes while augmenting human decision making. One of the key characteristics of an AI system is the ability to converse in a natural language, to comprehend conversation and interact with intelligent responses.

Dive in a bit, and you will realize that most of these companies are putting Alexa and Google Home as their front-end, calling it AI just because they can talk! These human-like interface devices are just empty shells, and they need a robust AI engine working behind the scenes for these devices to exhibit the true behavior of AI.

Let’s get into the real nut and bolts to understand what it takes to build an AI engine. Below is a picture that will illustrate high-level steps involved in building a true Enterprise Class Artificial Intelligence-Based Conversational Engine that can demonstrate real business value. 

diwo cognitive computing platform

Life cycle for an Artificial Intelligence-Based Conversational Interface.

This should give a fair understanding of the high-level steps to build an artificial intelligence application, involving neural networks, semantics, NLP, domain ontologies, etc.

That said, we are rapidly replacing the traditional era of programmatic computing with a new age of ‘Cognitive Computing.’ This is made possible through innovations around natural language programming and neural networks. In a nutshell, it means that we don’t need to be explicitly coding everything to accomplish a certain task, and are marching towards systems that can learn and augment human intelligence.

AI is the new UI—experience above all. Artificial intelligence is coming of age, tackling problems both big and small, by making interactions simple and smart. AI is about to become a new user interface and change the way we transact and interact with our systems. It’s much deeper and more revolutionary than the superficial addition of AI connectors.