In just a few years, Artificial Intelligence (AI) has evolved from an experimental technology reserved for innovators to a foundational driver of modern convenience retail. In his recent conversations on the Amazon Web Services (AWS) podcast and YouTube channel, PDI Energy and Retail COO Steve Antonakakis highlights this rapid AI transformation.
Steve shares lessons from joint PDI and AWS projects where AI is optimizing operations, unlocking new value from data, and empowering retailers to act with more precision and confidence. His perspective paints a clear picture: the earlier retailers adopt AI, the faster they’ll reap the advantages.
Unlocking value through better data and standardized operations
Convenience retailers generate enormous volumes of data every day, from POS transactions to fuel pricing, loyalty engagement, logistics, and even equipment tracking. Unfortunately, this data has historically been too scattered, inconsistent, and difficult to act on.
Solving that data problem is critical before deploying AI. Steve points out that retailers must first understand where AI can help and ensure that their underlying data foundation is solid:
“Decide what is going to bring you the most value and then go figure out the tooling. The tooling and technology are almost the easiest part, because there’s not a lot initially you need to do to get high value quickly. The first thing would be, can you get your data?”
To accelerate this effort, AWS and PDI have created an “affinity matrix,” a heat map that identifies internal work processes with the highest potential for generative AI impact:
“We said, what functions have the most repeated work we could potentially automate? What teams have really complex needle-in-a-haystack situations where there’s just a lot of data? And that showed us what to go after first.”
This structured, data-driven approach has allowed the team to move rapidly from exploration and modelling to execution:
“Within a week of starting—honestly more like three or four days—we went from zero to a practical AI application being used by our marketing team. It blew our mind.”
By combining AWS technologies like Bedrock, Kendra, and S3 with PDI’s extensive data, the team has quickly demonstrated how AI can streamline operations and reduce manual effort.
Using predictive intelligence for smarter store and fuel operations
AI also brings new predictive capabilities that can help retailers anticipate needs before issues arise. Whether it’s equipment maintenance, fuel demand forecasting, dispatch optimization, or inventory replenishment, the impact is potentially broad and significant. Steve highlights several high-value use cases underway within PDI:
“We’re already in the optimization space for fuel pricing. LLMs will allow us to bring that into more advanced approaches—like deep reinforcement learning—so we can converge and train faster and get results quicker.”
In transportation and logistics, the opportunity is equally promising. Predictive intelligence not only reduces operating costs but also minimizes downtime, improves service levels, and enhances the overall consumer experience. Steve notes:
“Making sure all the trucks on the road are delivering fuel on time, minimizing time on the road, and saving fuel—those are big opportunities.”
Navigating compliance without slowing innovation
As a global company operating in sectors considered to be critical infrastructure, PDI has to carefully balance innovation and regulatory discipline. Data privacy remains a top priority, and PDI is actively engaging with emerging regulatory bodies in Europe to strengthen oversight with the burgeoning use of AI.
The heightened exposure is why PDI is revisiting its full governance framework, with added focus on ethical model selection. Steve explains:
“Why is generative AI introducing new risk to data privacy, compliance, and ethics? It’s ubiquitous…you can inadvertently push information that isn’t yours to push. It’s important that any tooling reflects our values.”
Despite the complexity and the need for careful monitoring and assessment, Steve says guardrails don’t necessarily have to slow progress:
“We believe we can still move quickly as long as everything is checked. That’s how we stay quick and nimble.”
Moving toward a more connected convenience ecosystem
According to Steve, the next wave in AI transformation is all about combining search capabilities with LLMs to generate fast, accurate answers from structured data.
For example, in ERP and operations workflows, AI promises to eliminate manual steps that previously couldn’t be automated. In fuel pricing, blending existing models with LLMs—even deep reinforcement learning—will speed up staff training and improve forecasting. For logistics, AI can optimize routing so trucks complete deliveries on time while reducing fuel use.
Get More Insights on AI in Convenience
Discover the top trends and updates on how AI is impacting the fuel and convenience retail sector.