by Clint Boulton

5 machine learning success stories: An inside look

Aug 10, 20208 mins
Artificial IntelligenceIT LeadershipIT Strategy

IT leaders share how they are using artificial intelligence and machine learning to generate business insights.

cio ts ai ml by just super getty images 2400x1600
Credit: Just_Super / Getty Images

Artificial intelligence and machine learning (ML) are gaining significant traction in the enterprise, with organizations increasingly harnessing the technologies to better anticipate customers’ preferences and to bolster business operations.

Spending on AI systems will top $97.9 billion in 2023, nearly triple the $37.5 billion spent through 2019, according to IDC. And 87 percent of 950 organizations surveyed have deployed AI pilots or launched limited use cases into production, according to Capgemini research published in June.

Yet the COVID-19 outbreak presents a new challenge for AI, as many organizations that rely on historical data to shape their algorithms have seen their models skew since March. This “data drift” phenomenon makes it difficult for companies to rely on their existing models, says Jerry Kurtz, Capgemini’s executive vice president of insights and data. For example, models will likely change significantly for a company trying to predict maintenance intervals for jet engines, the use of which has fallen off in recent months. Ditto for retailers that have watched sales decline in recent months.

“There is a good percentage of cases where certain data changed so rapidly that history is no longer a good predictor,” Kurtz tells “Companies will have to revisit their algorithms because they never assumed the variables would change.”

In the face of such challenges, CIOs who are implementing AI and ML discuss their work.

Health insurer tabs AI to boost business outcomes

Anthem has implemented AI and ML solutions for tasks ranging from anticipating the trajectory of a patient’s health to mitigating disputes over its services, according to Rajeev Ronanki, chief digital officer of the health insurer.

By analyzing years of health-care data generated by patients with chronic conditions, such as diabetes or heart issues, and cross-referencing it against patients with similar conditions, or their “digital twins,” Anthem can anticipate the likely outcome of treatment.

AI also helps Anthem monitor the progress of claims and other services to detect potential customer issues with benefits claim adjudication and other services. If Anthem identifies a looming discrepancy, its customer service team will reach out to a provider or patient to explain the rationale. Such proactive outreach, Ronanki says, is critical in preventing volatile encounters. To do this, Anthem analyzes historical data collected from millions of calls in which customers expressed dissatisfaction with the company’s services. The AI generates scores signifying the likelihood that customers could escalate complaints. 

Among the AI vendors Anthem has partnered with is CognitiveScale, which Ronanki credits with helping the company democratize AI and reduce the cost of development of high-impact-use cases. CognitiveScale, which specializes in facilitating explainable AI, helps Anthem “explain our decisions, and provide the context that we can.”

AI capabilities and skillsets are embedded across every business line at Anthem, with applications developed by cross-functional teams for the express purpose of simplifying the health-care experience to make it more “personalized, productive and proactive,” Ronanki says. 

Shipping company shores up parcel processing with ML

Pitney Bowes, a 100-year-old provider of office shipping and mailing services, has been working with AI and ML tools extensively for the past 8 years, Chief Innovation Officer James Fairweather tells The company is currently using ML software to predict when its mail and parcel stations, comprising an Android tablet and integrated printer, are likely to fail. If the ML software, which communicates directly with the connected stations, senses a potential malfunction, it schedules a field service technician to work on the machines.

Fixing problems before the machines fail is critical for reducing downtime for shipping parcels, Fairweather says. And since the ML software has gotten so good at anticipating problems over time, Pitney Bowes schedules the service session neatly into the field service management system. “It provides a great experience for the customer,” Fairweather says.

With the consumer experience around delivery becoming so critical in a world where one-day shipping is becoming increasingly common, Pitney Bowes also uses ML algorithms to optimize shipping return volumes by monitoring parcels’ routes to identify sequential anomalies in processing. For instance, the algorithm will flag a parcel that typically gets scanned every 4 hours along its route but misses the second scanning window, Fairweather says. “We built a data science model based on the norms of these activities to predict anomalies in processing,” Fairweather explains.

Cranberry producer juices up operations with ML

Before Ocean Spray embarked on its AI and ML journey, the maker of cranberry, grapefruit and other juices had to clean up years of data it had collected. The company executed a master data management strategy to improve the uniformity and accuracy of its information assets generated by its business units and customers, Chief Digital and Technology Officer Jamie Head tells

Ocean Spray is using ML to comb through the past three years of historical data to gauge trends in sales uplifts, as well as analyzing patterns of competitors’ promotions to address any season gaps it might have, Head says. Head’s team is working with ML startup Visual Fabric to help understand how it can better generate insights from its trace spend, to “drive the business forward,” Head says. The IT group shares these insights with the sales team to help them refine their approach to market.

Ocean Spray is also exploring how to use ML to improve the quality of its cranberry yields by analyzing the colors, size and other variables, including soil and climate conditions for its farming partners in Canada, Massachusetts, New Jersey, Wisconsin and Chile, among other regions.

Machine maker manages sales with virtual assistant

Honeywell’s sales staff uses AI software to help prioritize meetings and manage leads that help them reel in customers for the company’s avionics systems, construction vehicles and other industrial machines.

The software, a virtual assistant built by, pulls information from Honeywell’s Microsoft Office 365 and Salesforce systems, according to Patrick Hogan, the industrial manufacturer’s vice president of commercial excellence. Using their smartphones, staff can speak to or text the Assistant to check whether they are on track to meet their sales goals, and view metrics on how customers have interacted with their business proposals.

When a sales staffer finishes a meeting, the Assistant will ask them what next step they plan to make. The Assistant also “nudges” users with notifications to follow up on opportunities that may be growing stale. “It helps you stay on top of your territory,” Hogan says, adding that the tool learns more about each sales staffer’s workflows and preferences with each use.

Assistant has had such a positive net impact on Honeywell’s sales funnels, including more face-to-face meetings and increased dollars per seller, sales conversions and yield rates, that he is actively urging more of the company’s 9,500 staff to use the tool.

AI boosts business services personalization

Office Depot is investing in ML capabilities to generate insights about its customer preferences and better recommend products, according to CIO Todd Hale.

The analytics effort comes as the $11 billion company seeks to expand its business services division, including its CompuCom tech services unit, while reducing its reliance on office supply sales. B2B sales fuel more than 60 percent of Office Depot’s revenues. The company uses advanced AI/ML techniques such as XGBoost and random forest to segment its customers into personas, and to predict churn, customer lifetime value and product affinity.

“In e-commerce, we utilize the deep learning power of Analytics Zoo on Apache Spark and BigDL to provide real-time, user-based product recommendations and develop cross-sell and up-sell models,” Hale says. Ideally, this will help Office Depot create “tailored products and services,” he adds.