New Research Identifies Scale and Automation as Common Keys to Successful AI

A new survey report commissioned by Wallaroo.AI, the leader in scaling production machine learning (ML) from the cloud to the edge, identifies the common characteristics among organizations that have found success in their artificial intelligence (AI) initiatives. The survey, “Lessons from Leading Edge: Machine Learning Best Practices and Warnings from Chief Data Officers,” provides a clear picture of how leading-edge organizations find business value from ML, how they plan to invest in the near term and the challenges they expect to face in achieving their ambitious goals for getting new ML initiatives into production. Having found a successful formula, most of those surveyed plan to dramatically increase their spend on ML and use of ML models in the near term.

CIO INFLUENCE News: Ivanti Reinforces Commitment to Cybersecurity and Customers with New Product and Engineering Leadership Hires

“The significance of this report cannot be overstated”

“Leading edge ML enterprises have a number of lessons to teach other organizations embarking on their own ML production journeys,” said Vid Jain, founder and CEO of Wallaroo.AI. “That’s why we commissioned this research to help understand how successful organizations have generated real business value from ML. Equally important, we wanted to learn how they avoided the pitfalls that prevent most ML projects from reaching production let alone having a significant impact on the bottom line.”

Finding Common ML Initiative Success Factors

The survey, conducted by NewtonX, asked Chief Data Officers, Chief Analytics Officers and leaders responsible for AI business outcomes in U.S. private enterprises about their AI initiatives. Despite prior industry research indicating that 90% of AI initiatives fail to produce substantial ROI and roughly half never leave the prototype stage, the overwhelming majority of respondents to this survey (92%) find business value from their models in production and two thirds (66%) feel their models have delivered results that are outstanding or exceed expectations.

Common use cases for AI among these leading-edge organizations include personalizing the customer experience, fraud detection, optimizing sales and marketing and improving real-time decision making. Their success of this group offers a basic roadmap that other organizations should consider when developing their own best practices, including:

• Approach: A majority of responding organizations have a robust, defined approach and a dedicated team for monitoring ML models in production. In fact among larger enterprises, 71% have at least 100 people working in ML while over half have more than 250. At the same time, 80% of respondents rely on automation in the ML testing and monitoring process.
• Impacts of Investment. Leading-edge companies that are already finding value in ML have invested heavily and plan to increase investments. One quarter of respondents spend $25 million per fiscal year on ML; two thirds are currently spending more than $10 million per year; 84% are spending more than $5 million. Having found success, most firms plan to increase their expenditures on ML ; in the next 3 years, two thirds of organizations in the survey expect to at least double their ML spend and 34% plan on at least quadrupling it. However, most organizations do not have the ability to invest time or money at this level, further highlighting the need to seek automated solutions and other ways to increase efficiency and scalability.
• Scale: Leading edge organizations in the survey believe that scaling is essential for generating ROI from ML. In fact, 36% of respondents expect a 10x expansion in their use of ML models over the next 3 years; nearly all respondents to the survey plan to scale their use of ML more than fivefold over the next 3 years.

CIO INFLUENCE News: Mavenir and i2i Systems Partner to Accelerate the Adoption of Open RAN in Türkiye

ML Ambitions vs. Challenges

Leading edge firms have not achieved success in ML without hitting some bumps in the road. While some of these obstacles are being addressed as the market matures and new tools and platforms come online, others will continue to have an impact.

These include:

• The need to build production frameworks: Many organizations have had to build their own frameworks from point solutions, driving up the costs associated with software, development talent, and external consultants. These do-it-yourself frameworks also tend to be complex and difficult to manage, leading to downstream costs. But as the market has matured, organizations can now turn to turn-key frameworks.
• Lack of talent: The lack of ML experts is the single largest issue preventing organizations from realizing their ML ambitions, cited by two thirds of respondents. If large, well-resourced organizations are having trouble finding and retaining talent, that has obvious implications for smaller organizations that want to adopt ML.
• Time to value: The complexity and cost of implementing ML projects, particularly for organizations that have not optimized their platforms, is a burden that not every organization can tolerate, given the high numbers of respondents who cited cost and complexity as challenges to success.. This is placing an even greater emphasis on finding platforms that are simpler to implement, easier to manage and provide automation.

Regardless of whether next-wave enterprises decide to build their own operations platform for ML production, it’s critical that organizations of all sizes are able to deploy, scale, monitor, and optimize all stages of the ML production process efficiently.

“The significance of this report cannot be overstated,” said Patiwat Panurach, VP of Strategic Insights and Analytics at NewtonX. “The implications of such rapid growth and expansion in ML are profound, placing intense pressure on Chief Data Officers. In a landscape where knowledge is power, this report, which is based on the expertise of the enterprise leaders shaping the future of machine learning, will enrich the strategies for other Chief Data Officers, and help them build a robust foundation for the ML-powered future.”

Among the notable takeaways from the report, while AI is still in its early growth phase, there is enough value to enterprises to make AI a good investment, particularly for industries that can take a competitive advantage form personalizing customer experience, fraud detection, optimizing sales and marketing, and improving real-time decision making.