Reading Time: 5 minutes

This is the year an overwhelming majority of enterprises are seeing value from their investments in artificial intelligence (AI) and machine learning (ML). According to a 2022 survey by NewVantage Partners, 92 percent are now achieving returns on their investment, up a stunning 44 percent from 2017. More than 26 percent of those surveyed have these systems in widespread production, double that of 2021. 

Even though ML is a technical discipline, ML goals ultimately must serve the enterprise. Since value will be defined by the business, the path to value from AI/ML requires business leaders in non-technical roles to help translate business challenges into ML use cases. These leaders also need to be aware of common ML use case risks as well as rewards. 

3 Common ML Use Cases

The power of ML is the power to scale data-driven automation and discovery. This translates into three important, common ML business use cases:

• Process Automation: For manufacturers, logistics and supply chain providers, and other industries reliant on robotic processes, ML predictive models bring value by enabling automated decision-making at scale. 
• Insights Discovery: Across industries such as healthcare, biopharma, manufacturing, and energy, ML can exponentially increase answer generation for “what if?” scenarios and can assist in predicting which options will yield the best outcomes.
• Anomaly Detection: When optimizing processes, identifying anomalies is key to success. ML can sift through the data “noise” at scale, enabling automated remediation and alerting for human intervention.

There are rewards, and risks, associated with each common use case that business leaders need to understand:

ML can help the enterprise across multiple dimensions and in industry-specific ways.

• More efficient maintenance. For example, manufacturers can improve equipment maintenance, processes, and product yield. 
• Optimized logistics. Logistics and supply chains will benefit from ML-driven automatic routing optimization, improved fuel efficiency, predictive maintenance alerts, and warehouse space optimization.
• Innovation at scale. Across various industries, ML can drive exponential increases in “what if?” answer generation enabling innovation at scale. 
• Faster time to action. Action automation can be improved when best actions are identified and iterated by ML systems.
• Optimization at scale. ML value in scaling anomaly detection is not just for manufacturing optimization. It’s a powerful tool for any industry where rapid, at-scale identification of anomalous data can help meet the need for fast automatic or human intervention toward optimization.

Without effective ML model operations and sufficient training data, those rewards can come with significant risks.

• Erosions in and degradation of performance over time. Loss of advantage can come from degradation in outcomes due to bias, and model drifts over time expose the enterprise to risk. For example, closed-loop algorithmic learning can amplify and entrench errors, and machine learning model drifts over time can result in suboptimal outcomes.
• Misdirections and missed directions. The quality of insights is dependent on the quantity and quality of data used to train ML systems, as well as risks from model drift and bias. Without partnership across data management and analytics teams, without robust model operations, ML can lead to misdirection.
• Salience gaps that consume resources with little benefit. ML is a powerful tool for scaling optimization, but not every anomalous signal matters. Without close partnerships between business stakeholders and data scientists, resources may be wasted on low-value optimization.

How Business Leaders Can Help Manage ML Reward and Risk

While it’s clear that ML can accelerate the enterprise, McKinsey notes there’s a high rate of ML value failure. A survey found “only about 15 percent of respondents have successfully scaled automation across multiple parts of the business. And only 36 percent of respondents said that ML algorithms had been deployed beyond the pilot stage.” Without working partnerships between business stakeholders, analytics teams, and data science teams, your organization may join the ranks of those struggling. 

These three best practices will help business leaders in overall ML efforts so that it helps and doesn’t harm your business:

1. Focus on those metrics, and the data, that matter most. Without clear business goals and absent clarity on what metrics and data are needed to support those goals, even the best data scientists will be flying blind. Business leadership is uniquely positioned to provide that clarity and to help prioritize what to focus on first.
2. Ensure business stakeholders have a voice in analytics initiatives. Whether an organization-wide Center of Excellence or a small working group piloting analytics projects, the outcomes will be better for the business if there’s meaningful participation by business stakeholders. Make sure your IT and analytics partners have your attention.
3. Invest in data literacy and data management. Ultimately the value derived from ML will depend on how well the model operations and data management processes within your organization can support it and how your non-technical people can use ML insights. Too often, enterprises relegate ML to a “technical expertise silo.” This can doom your ML projects by failing to provide the quality training data ML systems require and failing to connect to the business value needed.

Aeroporti di Roma Uses Machine Learning to Delight Customers

Aeroporti di Roma wanted to create the best customer experience for its guests both on planes and at the terminal. Deciding where to add new stores, restrooms, water fountains, and more was not just a matter of space. For the best flow and organization, optimal placement of these amenities would ensure maximum passenger comfort and seamless transitions—from the minute they step into the airport to when they board their plane. TIBCO platform capabilities enabled Aeroporti di Roma to use ML for “what if” scenarios, speeding its exploration of how different layouts influence passenger flow around the terminal.

According to Floriana Chiarello, Head of Demand Management at Aeroporti di Roma, “We are focusing on passenger flow analysis to better understand customer needs and how to improve the time they spend in various areas. We use TIBCO to develop passenger heat maps and bubble maps that show volume and density, as well as typical passenger flow paths.”

Aeroporti di Roma is now the leading airport in its class. It was named the best airport in Europe for the last four consecutive years by Airports Council International (an accolade never before bestowed upon a single airport over such a span of time) and rose from a 3.31 to 4.47 out of 5 in the group’s “Airport Service Quality” index, a massive 35 percent increase in the coveted rankings.

ML is a key ingredient in enterprise digital transformation. Business leaders must help their technical experts use it in ways that will matter most for business outcomes.

Get Started with TIBCO

Now that you understand three key ways ML can help (or harm) your business, it’s time to take the next steps in learning. Start by reaching out to your data science and analytics teams to support them in getting ML value.

And be sure to check out the latest information from O’Reilly on ModelOps: Ten Things to Know About ModelOps to be Successful.