How a 180-Year-Old Postal Service Reduces Customer Calls by 33% and Improves Warehouse Safety with Predictive and Generative AI

DataRobot ticked all the boxes for data science capabilities. We now have champion/challenger comparisons, health metrics, and hands-off model monitoring.

AI for Better Customer Experiences and Supply Chain Optimization 

Like any postal service today, NZ Post is navigating tremendous change as mail volumes decline and parcel volumes increase. Last holiday season alone, NZ Post delivered nearly 17 million parcels across New Zealand.

As parcel quantities grew, so did the number of customers reaching out to the call center with questions or concerns, requiring more from their support teams. Their warehouse teams also needed greater levels of support to ensure that packages were delivered securely, on time, and with the highest possible safety standards. 

To handle these challenges and more, the 180-year-old NZ Post turned to AI to help manage their customer experience and rethink their supply chain management and warehouse safety practices. 

In the past, the team relied on open-source coding languages and basic AI tools for predictive analytics but found them too limiting.

“We wanted something more robust and more professional,” explained Tim Reed, Head of Data Science and Analytics.

A Complete Platform for Predictive AI and Generative AI 

NZ Post chose DataRobot for its complete AI lifecycle capabilities, image classification, and industry-leading governance for both predictive and generative AI models. It also integrates with the company’s broader technology ecosystem, including Snowflake within Amazon Web Services, Microsoft Power BI, and Azure OpenAI for generative AI. These integrations drastically reduce siloed workflows and ensuing inefficiencies, saving NZ Post valuable time and effort. 

“DataRobot ticked all the boxes for data science capabilities,” Reed said. “We now have critical champion/challenger comparisons, health metrics, and hands-off model monitoring.”

As data scientists ramped up, Reed encouraged them to apply DataRobot quickly to new and existing models to pinpoint the highest-performing and most accurate models. 

“Our team loves it,” Reed said. “The user-friendly interface made it incredibly easy for them to start experimenting and delivering results. 

Harnessing Predictive and Generative AI to Reduce Customer Call Volume by 33%, Supporting Cost Management

Customer service — whether it’s delivering packages on time or keeping recipients updated — is an essential part of NZ Post’s commitment to their customers. If a customer’s concern is escalated to the helpline, there’s a financial and reputational cost to NZ Post. Therefore, reducing call volumes was the top priority for NZ Post as they embarked on their AI journey. 

Using the DataRobot AI Platform, the team can predict which customers are likely to call within the next few days and their most probable reason for calling. These predictions enable NZ Post to get ahead of customer questions and address their concerns. 

“Those top predictive explanations act as the prompts for a generative AI LLM to create emails with a theme and context that relates to a customer,” Reed said. “It’s not specific to one person but segmented based on predictions.” 

NZ Post automates the entire process of segmenting and emailing customers with guidance on a particular question. Ideally, their proactive communications answer a customer’s question, preventing a call.

Even in a small trial, these complementary use cases for predictive and generative AI produced impressive results — reducing call volume by 33%, according to NZ Post.

With DataRobot, NZ Post manages predictive and generative AI models in a single platform, unifying their AI landscape, teams, and workflows for full visibility and oversight at scale.

“With DataRobot’s generative AI monitoring, we can pick and choose which LLM to use and decide which gives us the best answer for our problem,” Reed said. “It’s complete with the metrics that relate to that level of accuracy for toxicity, bias, etc.”

The team further improved their customer experience by building a model to segment customers for more personalized banners on their website. By refining customer communications, NZ Post noted they were able to reduce visits to their call center contact page by 18% and deliver an 8% reduction in those calling after visiting that page.

Merging predictive and generative AI also helps customers narrow down the right tariff codes when sending parcels internationally. Out of 16,000 potential tariff codes, they find the closest matches – reducing frustration and elevating the customer experience.

Driving Privacy and Warehouse Safety with Image Classification

Perhaps the most exciting results from NZ Post’s partnership with DataRobot, however, have stemmed from leveraging image classification to improve consumer privacy and warehouse operations. DataRobot Visual AI lets NZ Post incorporate image data easily into its models alongside tabular and text-based data types. 

When delivery people bring parcels to recipients, they take photos as proof of delivery. However, due to privacy laws, they can’t retain all personal information present in the photos. Image classification within DataRobot allows NZ Post to blur identifying parts of a photo to maintain privacy while assuring customers that their packages have been safely delivered. 

The postal service also analyzes closed-circuit TV footage from its warehouses for potential safety incidents. For example, if a forklift and a human get too close, it triggers a proximity alert. Then, they can flag that incident in a specially-created dashboard. The near real-time issue-tracking allows NZ Post to build continuous awareness of best practices while adapting to the surge in parcel deliveries, allowing their workers to operate efficiently and safely. 

“Image classification has been one of the main differentiators for DataRobot,” Reed said. “We’ve presented our warehouse safety use case in a few places and always get ‘oohs’ and ‘aahs’ because it’s pretty cool.”

Being able to integrate multiple data formats into their AI programs helps NZ Post maximize the value of their data and improve operations across supply chain operations, the customer experience, and digital transformation. 

Industry-Leading Governance and Observability Inspire AI Confidence 

Though their early AI projects have already resulted in significant savings and improvements, Reed and the NZ Post team are most inspired by DataRobot’s industry-leading governance capabilities. DataRobot’s unified oversight and governance of their predictive and generative AI means that NZ Post can further expand their AI program while remaining confident that their applications are secure, compliant, and accurate.

DataRobot gives us reassurance that we are accessing generative AI through a well-governed and secure environment.