AI in Business: The Transfer of Expertise from Internet Companies to the Enterprise
The enterprise has started to integrate AI and ML into its operations, but not nearly to the extent that many internet businesses have. Help from these companies could be the key to enterprise AI adoption.
Hyperscale internet companies have leapfrogged several levels of machine learning with increasing automation in data processing and modeling sophistication since 2015. The enterprise, with a few exceptions, has been lagging in adoption of artificial intelligence but sees, in internet companies, partners who can help it to catch up.
The prospective enterprise users of machine learning have a long way to go to match the talent pools, computing prowess, scale, and the data volumes for training algorithms that internet companies have accumulated, especially over the last four years. In many verticals of the enterprise, the business processes have not been digitally transformed for the automation of data processing and the instant execution of business decisions based on insights gained from artificial intelligence. Moreover, several of the verticals do not yet have well-defined use cases that lend themselves to the profitable execution of artificial intelligence. (For more on AI in business, see Overcoming IT Service Management Change Management Woes With the Power of AI.)
Adoption of Artificial Intelligence in Business
Adoption of artificial intelligence in business is at an early stage, especially when we consider its sophisticated users who have gone beyond exploration and pilots to a stage where they gain business value from its usage. O’Reilly, a technology media company, found in its 2018 survey, “The State of Machine Learning Adoption in the Enterprise,” that sophisticated users were only 15% of the total enterprise users worldwide and 18% in North America.
External sources of expertise and learning play a significant role in aiding business users to catch up with the state-of-art in machine learning, especially for advanced AI techniques. A 2018 survey by Deloitte found 59% of the enterprise buyers acquire AI expertise from enterprise software companies with AI capabilities, 53% co-develop it with partners, 49% acquire it from cloud AI companies, and 39% crowdsource it from sites like GitHub. Cloud AI companies provide AI as a service, which saves on the cost of infrastructure and talent development on-premise.
For advanced AI development, cloud companies are a more important source of expertise. Thirty-nine percent of the business respondents showed a preference for cloud companies as a source of advanced AI compared to 15% for on-premise software. AI as a service has grown at a brisk rate of 48%.
Adoption of Artificial Intelligence in Verticals
We spoke to Aditya Kaul, research director at Tractica, an industry analyst firm focused on artificial intelligence and robotics. Kaul has been investigating the adoption of artificial intelligence in 30 verticals for over 300 use cases in businesses across the world. “Telecommunications and financial services have been the leaders in AI adoption, and they started early with more rudimentary statistical techniques going back as far back as the 1980s,” Kaul told us. “Adoption in retail, automotive and healthcare has surged in more recent times while the majority of the enterprise remains at an early stage of adoption,” he added, “Horizontal business services such as CRM, supply chain, and HR have expanded the adoption of AI rapidly as its predictive capabilities help in identifying prospects, consumer demand trends, and talented employees.”
“Monitoring, synchronization, and optimization of complex and heterogeneous software-defined networks is a critical use case in the telecom sector,” Kaul surmised. “Voice-assistants in cars have surged in the automotive sector with an increasing accent on the in-car personalization of services,” he noted. He also informed us that “The banking sector is deploying artificial intelligence for customer service including chatbots as they face intense competition from smaller internet banks, apart from using it for fraud detection, loan analysis, and other backend operations.”
While the healthcare sector has enormous potential, it had lagged until recently due to regulatory barriers to using its data. “Several venture-backed start-ups have now focused on machine learning in clinical trials to speed up drug discovery,” Kaul revealed.
Retail stores have accelerated investments in machine learning as they achieve mastery in predicting demand and supply accurately. German retailer Otto cut returns by more than 2 million items a year and excess stock by 20% using deep learning algorithms to predict what customers will buy, according to a research report by McKinsey. Its AI engine now autonomously orders 200,000 items a month because it can forecast what Otto will sell over the next 30 days with 90% accuracy. (Not sure how AI would fit in with your company? Check out 5 Ways Companies May Want to Consider Using AI.)
Partnership with Cloud AI Companies
Hyperscale cloud AI companies have been willing to partner with enterprise customers to advance their artificial intelligence skills, but they are uncertain about the ways to collaborate with enterprise software companies who are indispensable for backend plumbing. “Cloud companies have been generous to enterprise customers with their freebies including free cloud time, consulting, and training resources,” Kaul observed.
Since cloud AI companies like Google have made a quick transition from hand-engineered algorithms in 2015 to deep learning in 2016 and lately more advanced algorithms like reinforcement learning, they are able to counsel early adopters on how to make progress in their journey to AI learning maturity.
“The costs of AI are also dropping as we see increased availability of pre-trained models, labeled datasets and a general reduction in cloud AI pricing,” Kaul explained. “Concurrently, the time for data processing, ingestion, data preparation, and labeling, which accounts for 90% of the effort, has been shortened with techniques like AutoML which automates these processes,” he added. Nvidia, a partner of hyperscale cloud AI companies, has repackaged its GPUs (graphical processing units) for the enterprise. “Nvidia has repositioned to target data science and analytics use cases in the enterprise which speeds up the training of large analytical models compared to CPUs (central processing units),” Kaul explained.
Enterprise software companies will have to find a way to accommodate cloud AI companies, especially as they bring new capabilities to the market which become a part of the fabric of enterprise business. “Functions like chatbots and computer vision capabilities for image recognition are enabled by deep learning which extends the value that AI brings,” Kaul asserted. “Software itself is not hardcoded anymore but adapts to needs of data and analytics,” he added. There is, as yet, insufficient evidence to show that enterprise software companies, with a few exceptions like Microsoft, can catch up with cloud AI companies in algorithms. By all indications, the new terms of engagement between cloud AI companies and enterprise software companies, however, have not been resolved yet.
Machine learning will reinvent the enterprise as it redefines enterprise software itself. The enterprise will adapt faster to the external business environment with the automation of data processing and faster execution of business decisions based on insights gained from algorithms that shorten the time to learn from data. Enterprise software will evolve and reconfigure more often to keep pace with algorithms.
More from AltaML
- What are some of the dangers of using machine learning impulsively without a business plan?
- Why is so much of machine learning behind the scenes - out of sight of the common user?
- What is 'precision and recall' in machine learning?
- Why are GPUs important for deep learning?
- What’s a simple way to describe bias and variance in machine learning?
- What are some of the main benefits of ensemble learning?
- Will machine learning make doctors obsolete?
- What's the difference between machine learning and data mining?
- How does a weighted or probabalistic approach help AI to move beyond a purely rules-based or deterministic approach?
- Why is data visualization useful for machine learning algorithms?
- What are some of the foundational ways that career pros stand out in machine learning?
- How is AI technology going to affect the workplace in the near future?
- Why are some companies contemplating adding 'human feedback controls' to modern AI systems?
- What does 'connectionism' mean for business AI?