Do you as manager struggle with production planning and scheduling? As a bank employee, what if you could better predict potential credit and loan defaulters? What if the work that takes hours to get done by 50 employees, gets done within seconds by 1 person? What if your homes could ‘sense’ something being wrong such as a gas leak or unwanted intrusion and take preventive and precautionary actions unsupervised?
Enabling all this is the promise of AI
Today, AI is being used in a wide range of applications such as driverless cars, autonomous helicopters, virtual assistants and face recognition. Do these applications imply that AI is only for high tech firms such as Google, Apple, Amazon and Facebook? In this article we try to establish how AI is a notch ahead of Analytics and pitfalls that a regular enterprise can avoid in their AI adoption journey.
What exactly is AI?
Imagine a customer service application. Earlier, a customer interacted with a company through a phone and all the information about the customer’s likes and dislikes depended on data that was entered in the system by the customer service agent. If the system captured too much information, it would reduce the productivity of the call center and impact the customer experience negatively. However, a sparse information capture would reduce the richness of data and consequently the quality insights.
Simply put, AI is a decision-making technology. What is new though, is that apart from structured data that came from RDBMS, AI can also extract information from content such as images, videos, text and speech.
With AI, we can automatically analyze not just customer needs but her emotions as well. The interventions predicted by the AI application help us reduce the customer pain points precisely. This means that an enterprise would have higher number of loyal customers.
How is it different from Analytics?
In traditional analytics we needed a human expert to code the decision-making rules. On the other hand, an AI program extracts patterns from the data automatically and uses it for decision making. This means that decision-making rules are changed by the software itself with introduction of more and more data, constraints and machine generated insights without human intervention. This may seem like an abstract idea, but it can be crucial in industries such as fashion where demands are highly irregular and eclectic. Similarly, fraudsters often come up with ingenious ways to game software and the humans working on them. AI has the potential to not only identify potential threats but also prevent damage from occurring. Even more interesting is the fact that with automatic pattern recognition, over time, the programs become faster and more accurate.
What are the risks involved?
The biggest risk is the hype around AI. It creates an atmosphere of extreme hope or that of extreme fear. For example, we often read how AI has the potential to cause massive job losses. But most of us are unaware that even state of the art in AI today is no match for sheer human capabilities.
At a more operational level, two biggest risks are:
- Starting an enterprise AI initiative without a proper data strategy. After all, AI is singularly dependent on the quality of data. A thoughtful data strategy would avoid injecting human biases in the AI applications.
- Creating a mish-mash of AI applications. This can accumulate technical debt that needs to be repaid later in terms of re-architecting and re-factoring.
Seldom has there been a technology that has been as polarizing as AI and Machine Learning. Usually any new technology is scoffed at by terming it unnecessary and inconvenient. It is written off by terming current prevalent methods as more than adequate. Things are nowhere as bleak as they are being made out to be. It is in the best interest of businesses to embrace AI and Machine Learning tools.