(This is part one of a two part series describing how future AI outcomes might be hindered by today’s database bottlenecks.)

A recent report by The McKinsey Institute demonstrated why it takes planning, patience and determination to truly realize and manifest the value of AI and Machine Learning in organizations. The report’s authors surveyed executives from roughly 800 companies in the technology, media and telecommunications domains, mapping these organizations’ AI journey over the last few years.

When they analyzed the results, they realized that only 10% of these companies, which were categorized as “leaders,” gained significant value from their AI investments. This value was not merely declarative or perceptive, but tangible and quantifiable: when comparing the operating revenue impact companies which were “leaders” (top 10%) registered during their AI journey, vs. the “laggard” companies (which were the bottom 60% of responders), the researchers found that the leaders outperformed the laggards by a factor of 3.4X.

They then dug deeper into the results and analyzed the specific journeys that companies from each ‘camp’ went through when they were integrating AI into their business. The researchers concluded that in order for an organization to truly utilize AI, the organizations need to look at AI as a long-term journey. This requires that they first understand what AI can potentially produce for the organization, but also adopt the mindset that this journey will probably include a lot of failures and frustrations. Yet these should not signify the results but rather the challenges on the way to the top.

The most important insight researchers gathered from analyzing the “leader” companies was that these organizations invested the time needed to go through the long processes of training the AI and ML tools using the needed data (which essentially means as much data as possible, especially when it comes to complex algorithms and use cases), and then worked step-by-step, focusing on just one or two areas within their business and trying to address multiple business cases at the same time.

By not cutting corners, by preferring patience and accuracy over speed and vanity results, and by focusing on specific domains and devoting their entire attention to them rather than multitasking, these organizations not only created AI-based processes which could handle more sophisticated use cases in these domains – they also created AI-based processes that were experts in these specific domains, because they had access to the relevant data and the time needed to teach themselves until they became experts.

Telecoms Reap the Benefits of AI, But for How Long?
As an organization which works with telecommunications service providers, we were interested in learning how they fared in this research. The results were interesting: the biggest gains in operating revenue by the “leaders” in the telecommunications camp – i.e. more than 3% – were from AI usage in customer care (67% of responders in that group), B2B sales (50%), and forecasting or demand planning (50%). Interestingly, churn prevention and up-sell/cross-sell impact by these companies added 1%-3% – still an improvement, though.

As someone embedded in the world of analytics and databases, I tried to connect these results with what I see today: the telecommunications service provider analytics bottleneck. This phenomenon is only increasing in severity, and is the result of the increasing amounts of data being created and stored by the telecoms from their users, networks, devices, etc., and the limited resources they have with which to effectively scale and rapidly analyze that data.

Antiquated systems or even more modern ones, which were built to handle smaller, less complicated, and less frequently updated databases, fall short of the tasks which today’s business needs require them to achieve.

So what does this mean for telecoms? Is there a way to train AI for farsighted, as opposed to short-sighted benefits? Check back for part 2 of this blog, where we will examine these questions and more.

To learn how SQream helps accelerate analytics for AI / ML models, read the datasheet.