From Resistance to Adoption – How Trailblazers Are Managing AI-Led Disruption
By Mukundha Madhavan
Uber, Netflix, FedEx — these names are ubiquitous, but what do a rideshare, streaming service, and shipping solutions provider have in common? Well, they were all early adopters of artificial intelligence (AI) and machine learning (ML) who were able to use that momentum to redefine their respective industries. Starting years ahead of much of the competition, these trailblazers were able to harness the right data, at the right time, to drive the biggest impact. The pioneering efforts of these brands have brought about a domino effect in some ways. Indeed, today 55% of digital disruptors have deployed AI/ML at scale, according to Bain & Co.
While it’s true these companies have been able to call upon vast amounts of data, the majority of organizations today are sitting on a data treasure chest of their own. The difference is that the likes of Uber and Netflix were quick to the punch because they were equipped to act on data in the moment and serve millions of customers in real time. Netflix tracks each user’s clicks to refine its recommendation algorithm, driving all user interactions and playing an influential role in determining its content. Uber, meanwhile, collects data from drivers, partners, and riders to refine its prediction software, which informs riders of wait times and suggests optimal routes for drivers. Then there is FedEx, who aggregates billions of package tracking events to constantly forecast outcomes, optimizing operations and providing shipment visibility to customers.
Another point in common between these powerhouses is that they all collect granular, time-stamped data. With immense amounts of detailed data, these companies can train models to reveal deep, contextual insights about influential factors and understand how events unfold over time.
Built for Speed and Scale
Given the massive volume of data required to handle the scale and speed required to build highly accurate models, giants like Uber, Netflix, and FedEx were able to get a leg up by accelerating the delivery of AI-powered apps, eliminating data transfer costs and accessing timely and actionable data. They built up their digital architecture to ensure their storage and handling of event datasets could be accessed instantaneously and with minimal latency.
They leveraged a distributed database that delivers the high availability, performance, and scale for high growth applications. This freed them from having to make compromises on data scalability, going to as much as hundreds of petabytes, while providing the throughput and availability required to improve applications and insights.
Furthermore, by virtue of a decentralized and replicated architecture, these giants were able to serve apps and models from the same datastore, so event data is available to compute features, at scale, without impacting applications. In addition, these frontrunners had the infrastructure to simplify access provisioning to apps and feature stores, ensuring the right data was accessible to ML models.
In addition, these companies also saw the importance of a modern streaming platform. This enabled connectivity to most event data sources, ensuring that events are made available in real-time, without complicated data engineering requirements. In addition, these companies were also able to undertake global replication, ensuring event data can fuel applications, feature engineering, and models with low latency.
Real-Time AI Is Possible for All
A key takeaway from the successes of these influential companies is that their achievements are not unattainable. In fact, now that the reason behind these wins are more accessible it becomes clear that most – if not all – businesses have the potential to deliver the kind of impact these companies have. That lies with harnessing the right unified data platform, so that ML initiatives have the right infrastructure and the right data.
When it comes down to it, it’s about empowering data engineers and data scientists to break down silos. While processes like feature engineering, model experimentation, training, and inference require many tools, it’s essential to align them with the same data foundation.
With the power of massive amounts of event data serving models and applications, the most successful AI-powered applications are able to differentiate and lead through constant improvement of customer experience and operations. Via their ability to serve millions of customers, and get smarter as they do, companies like Uber, NetFlix, and FedEx were able to leverage these apps and set new standards for the markets they’re in. While these capabilities were once out of reach for many, now organizations of all sizes are able to circumvent barriers like costs and skills to tap into real-time AI to applications at scale.
(The author is Mukundha Madhavan, APAC Tech lead, DataStax, and the views expressed in this article are his own)
