When Sankar Karuppasamy served as chief information officer at trading card manufacturer Topps, he reported to the chief financial officer. But after moving into the role of CIO at gum maker Bazooka, which was separated from Topps through a $700 million sale to buyout firm Apax Partners in 2023, the technologist began to report to the CEO, Tony Jacobs.
“There’s a lot of transformation going on, with the new owners supporting the investment,” says Karuppasamy of the role that technology plays in Bazooka’s future. The reporting structure shift reflects a broader trend across corporate America, as 65% of CIOs report to the corner office, up from 41% a decade ago, according to consulting giant Deloitte.
Karuppasamy has been deploying AI and machine learning quite broadly across Bazooka’s operations, helping sharpen demand forecasting, speed up marketing and packaging materials, piloting agentic AI in the company’s supply chain, and working on a project that transitions employees away from OpenAI’s ChatGPT in favor of Anthropic’s Claude. OpenAI and Anthropic are in the midst of an intense race that encompasses everything from who is developing the most advanced AI models to who can go public first.
Bazooka has seen its share of technological and economic change over its more than three-quarter-century history. After introducing its Bazooka chewing gum in 1947, the then-Brooklyn, New York-based company debuted Bazooka Joe Comics in 1953 and launched its “wearable” Ring Pop candy in 1977.
The advent of AI comes at a time when the food and beverage industry faces a number of challenges, with inflation-pressured consumers gravitating to private-label brands and overall consumption slipping as more Americans use GLP-1 drugs, which suppress appetites and lower calorie consumption. U.S. retail sales of confectionaries grew 1.5% in 2025, with unit sales down year over year in all categories (chocolate, non-chocolate candy, gum and mints) according to data from the National Confectioners Association.
Seasonality is another challenge for Bazooka. Candy sales spike around major holidays like Halloween, Valentine’s Day, and Christmas. That led Karuppasamy to invest in forecasting and scenario planning from the software vendor Anaplan, which he said utilizes machine learning to better forecast demand from retailers so that Bazooka can more precisely map out its manufacturing plans. Karuppasamy said forecasting accuracy has leapt to 90% from 60% before using machine learning in this manner.
Karuppasamy’s AI playbook is centered on three layers: productivity, changing unique workflows across different business functions, and an early effort to embrace autonomous, agentic AI. In the early days of the generative AI boom, Bazooka used ChatGPT for general productivity use cases.
But as the organization begins to evolve its thinking on AI away from generic use cases and more into developing AI tools for job-specific tasks and agentic AI, Karuppasamy said he was more impressed with what Claude has to offer. Still, he reserves the right to change his mind again. “Something else could pop up next year,” says Karuppasamy. “Because the space is so innovative.”
Bazooka offers employees foundational training to teach the basics of large language models, but while transitioning to Claude, it is also boosting department-level training that will focus on priority use cases and how to practically apply AI. The broad employee base has performance goals they need to hit for how they use AI, as do C-suite leaders when it comes to adoption and usage.
Some priority AI use cases that Karuppasamy has implemented across the business include facilitating faster recipe development, using AI to pull consumer insights from social media and retailer feedback, accelerating the production of marketing materials and ensuring all lingo that is used on its famous red-white-and-blue packaging adheres to local regulations, and enhancing software that’s used in Bazooka’s Pennsylvania manufacturing facility.
One technology advancement that Karuppasamy has resisted: robotics. “We don’t have a huge factory,” he says.
Karuppasamy is also spearheading a pilot program for agentic AI within the company’s supply chain. “Our first goal, before we run an agentic workflow, is to ensure we have a data-ready platform where the agent can work,” he says. Karuppasamy said Bazooka is keeping a close eye on the quality of the data, but also building the proper governance guardrails to ensure no data leaks out improperly.
He’s also chasing a proven outcome where agentic AI can show a tangible business outcome, with some priority use cases including freight management and tariff analysis. “We import a lot of our products from different countries, and it’s a big problem for us to continuously analyze it,” says Karuppasamy. AI, he hopes, can autonomously perform this work and show how tariffs will affect Bazooka’s finances.
John Kell
Send thoughts or suggestions to CIO Intelligence here.
