The impact of cognitive computing technologies—including artificial intelligence (AI) and machine learning (ML)—is increasingly being felt in data centers and database operations of all sizes, across all industries. A recent survey of 1,000 executives published by Infosys found that AI and related cognitive technologies are no longer just experiments conducted by computer or data scientists—they are part of a real-world technology wave that is already showing tangible business results. Eighty-six percent of respondents, in fact, reported that their organizations have “middle” or “late-stage” AI deployments, and view it as a key technology for future business initiatives going forward. Nearly three-fourths also stated that AI is already helping to transform the way their organizations do business.
Adoption is not moving at the same pace across industries, “but early adopters are working on areas such as connecting and engaging with end customers accessing products and services via mobile, TV, sensors, and call centers,” said Jai Ganesh, vice president and head for Mphasis NEXT Labs. “Adoption is also seen in areas such as supply chain for enhanced insights and experiences with regard to goods, products, and machines across their lifecycle usage. Offering location-based experiences, services, and payment processing by leveraging bionic sensors and hand-held devices is another area of interest among early adopters.”
Certain industries are outpacing others and moving on to enriching the user experience associated with AI-driven insights. “Some industries, such as financial services, retail, distribution, logistics, entertainment, and healthcare, have been pioneering research, development, and adoption of artificial intelligence and machine learning applications,” said Abay Radhakrishnan, CTO architect for Sungard Availability Services. “Natural language processing is increasingly needed to deal with data processing due to the global presence of the machine learning algorithms used for building applications.”
Words of Caution
While these reports of early cognitive computing success are encouraging, data executives and professionals can be forgiven for assuming that just about everyone is charging head-long into the cognitive computing space. While the technology is seeing success, it may take time and a good deal of implementation work until AI and ML begin delivering practical, on-the-ground differences. “The desire is high, but the action is slow,” said Ben Newton, director of operations marketing for Sumo Logic. “It turns out, AI and machine learning are hard. A big part of the problem is skillsets and knowledge on the one side, and false promises from vendors on the other.”
The current hype cycle around AI and machine learning “far exceeds the maturity or adoption of the technologies within enterprises,” said John Purrier, VP, software automation for CA Automic. “It looks as if we have the cart before the horse—the horse in this case being data that can be used to derive actionable insights.”
AI and ML are “hitting a wall,” concurred Manny Medina, CEO of Outreach, and former Amazon Web Services and Microsoft executive. He observed that while there is a lot of “noise” being created around AI, there is little value to show as of yet. ML “is still top down, is super data-hungry, and does only correlation-based assertions without bringing true business insight.”
Tellingly, the Infosys survey also revealed that the cognitive journey is a gradual one, and many companies are still just starting to realize its potential. A majority of respondents, 66%, indicated that their first AI projects are targeted at automating routine or inefficient processes.
Getting Data Houses in Order
To realize value in AI and machine learning, CA’s Purrier advised that enterprises get their data houses in order. “Enterprises need to first implement data readiness, which includes the ability to pull data from various silos and systems, ensure it is properly formatted—or cleaned—and stored in a canonical manner,” he advised. Medina added that businesses need to gather and process data in a very specific format: “It needs to be clean, to be labeled so that machines can learn, and to be closed-loop with inputs and outcomes. Businesses generally don’t have this available.” They can generate it, but very slowly, he added.