How AI is changing the nature of analytics

At its heart, artificial intelligence is an analytics tool. Its value comes from the ability to parse through enormous amounts of data, without direct human supervision, to identify patterns and anomalies that can then be put to use.

But since human-driven analytics have existed for centuries, long predating the modern computer age, how will this new generation of technology change the game? And how can organizations make sure they are getting their money’s worth once this technology is pushed into production environments?

A matter of context

The key element that AI brings to analytics is context, Oracle’s Joey Fitts and MIT research fellow Tom Davenport recently wrote in the Harvard Business Review. Under traditional analytics, the analyst was rarely an expert in the system or process being analyzed. They knew analytics, not marketing or sales or data networking. Their ultimate recommendations often lacked the context that can only come from broad knowledge and experience.

In an AI-driven framework, however, an algorithm can be trained to “understand” the thing it is analyzing and can then incorporate far more data at a much faster pace to deliver highly contextualized results. Ultimately, this is expected to push these powerful analytics tools to the people who require them so the analytics experts can devote their time to what they do best: crafting the models needed to make AI analytics faster and more accurate.

This need for context is best illustrated when applied to a common enterprise function, such as marketing. Arguably one of the most data-intensive disciplines in modern business, marketing is often subject to competing interpretations of the truth depending on the context in which data is presented.

AI excels at predictive analytics, the ability to spot future trends based on past and current data, according to Mike Kaput, chief content officer at Marketing AI Institute. This capability, of course, is like gold to a marketing team. At the same time, AI delivers prescriptive analytics — the ability to make recommendations based on predictive analyses. In both cases, today’s AI engines are capable of sifting through massive amounts of data to ensure these results are being presented within the full context of all available information, and they also have the ability to refine their algorithms to improve themselves using their own past analyses.

Learning the process

This ability to learn is one of the key differences between AI and simple automation. An automated system may still be able to parse a lot of data, provided it is structured properly and designed to address the specific needs for which the system was designed, according to analytics firm Avora. For instance, a simple reporting tool will update itself with new information over time, but it won’t be able to provide new insight into changing data unless someone builds a dashboard that allows it to do so.

Likewise, simple automation cannot answer general queries related to diminishing performance and other factors. This typically requires hours’ if not days’ worth of work by a data analyst, who more than likely will still only collate a limited amount of data. A properly trained AI engine, on the other hand, could produce results to multiple questions within minutes.

Perhaps the best way to view AI’s contribution to analytics is through one of the oldest analytical methods of all: the cost-benefit model. On the cost side, it requires a fairly sizeable upfront investment, provided you are building the underlying infrastructure from scratch. But this cost will amortize over time as output scales. On the benefit side, AI can crunch vastly more data than even an army of analysts could, and it can draw data from an untold number of sources to identify problems and/or opportunities that would otherwise remain hidden.

Ultimately, it will push analytics capabilities into the hands of knowledge workers who can best benefit from the insights tailored to their unique challenges, making the entire organization more efficient and productive.