Once or twice a decade, a buzz develops around a new way of doing business.

Today, that buzz is all about AI. But as with previous buzzes, there are questions to address. Is it evolutionary or revolutionary? Is it hype or is there substance? Can it help your business today or tomorrow? The answer to all these questions is “yes.”

The technology is evolutionary—AI has been around for decades through earlier versions of machine learning— but dramatic increases in processing power have allowed AI to reach revolutionary levels of performance including self-directed “reasoning” and decision-making. The newest AI, called Generative AI or GenAI, doesn’t just take data to predict, but to create something new—and continually learn and improve.

Is there hype? Absolutely. Some futurists predict that AI will dramatically transform business in ways that match or surpass the internet and will result in enormous financial gains while eliminating entire professions. Others, like Nobel Prize-winning MIT economist Daron Acemoglu, predict that AI will likely automate just 5% of tasks while adding only 1% to global GDP during this decade.

The promise ultimately lies in future applications. The capabilities that are just starting to be realized have the potential to create exciting new solutions to change the way companies interact with their customers—and drive notable increases in revenue and efficiency.

 

The Human Touch Is Vital

While there are predictions that AI could wipe out entire professions, no one is expecting the petroleum marketing industry to be one of them. In fact, the human touch will continue to be essential. Sophisticated AI goes beyond traditional data analysis, with elements of reasoning and decision making that make it even more important to keep humans in the loop, both during development and implementation, because there can be some odd results.

“You always have to keep the human in the loop,” said Kathleen Walch, the director of AI engagement and learning at the Project Management Institute. “Machines are good at looking at large amounts of data very quickly, and humans are not good at being able to spot all these different patterns and data, especially if you have terabyte sized data sets. Machines can do that, and then they say, ‘Hey human, why don’t you look over here and dig a little bit deeper into this?’”

Humans are critical for pairing solutions to specific market applications and removing the inconsistencies.

“You need someone who can talk to human beings who’ve been doing this and then can accurately convert it into models,” said Ishaan Grover, co-founder of Catalan.ai, a dynamic pricing platform that optimizes business performance in real time. Catalan is currently working with Price Advantage to enhance its industry-focused price modeling solutions. “By talking to our customers and having long conversations with them, we now know exactly how they’ve priced. So, then you have the whole picture that comes together.”

And once an enhanced AI solution is in place, active human monitoring is still needed.

“It’s not something you just flip on and say, ‘Okay, you’re going to be all set with this,’” said Dr. Adi Raz, head of data science at Titan Cloud, which provides fuel asset optimization solutions. “We must run the data and continually check it, fine tune it and then get it to a point where we all trust it. No artificial intelligence runs outside human interference. This product is meant to work with a dispatcher, for example. It can run by itself, but it really shouldn’t. The dispatcher will know something that the optimizer will not, such as if a truck suddenly becomes unavailable.”

 

AI and Predictive Applications

AI has plenty of real-life applications for the industry. “Say you’re considering predictive vehicle maintenance,” said Walch. “Now you do an oil change at 5,000 miles for every single vehicle and you have 100 vehicles in your fleet. It costs X amount of money, and you try to rotate it, but you find that that a lot of vehicles end up out of service at a certain time of the month and it’s causing issues. With AI you can begin to apply predictive maintenance—using sensor data to say which vehicles need to be serviced at what times and cycle them out so that you always have vehicles in rotation.”

AI can also help give retailers deeper insights into the broad universe that is impacted by pricing behaviors, demand or even the weather.

“If I move down on price today, many things can happen in the market,” said Catalan.ai’s Grover. “The competitors can decide to undercut me, and what do I do from there? Should I go further down? Should I go up? Will they go up? How the competitors are going to react is almost more important than anything else, and not just tomorrow but also in the future. What will I do if I got less volume because of this, what do I draw from it? AI can learn from this and eventually a new strategy is formed.”

Titan Cloud uses AI to predict how much fuel is demanded at a site on a given day, automatically assigning loads to trailers and drivers optimizing order assignments, optimizing delivery route planning and monitoring dynamic fuel demand over time.

Petrosoft used AI in its Predictive Competitor Fuel Price Pricing, even though it may not be seen as AI today, said Michael Munz, the company’s marketing operations manager. “That is AI-driven because it takes a ton of information and then quantifies it, digests it and gives it back to you in a manner that is efficient and effective. It eliminates a lot of human interaction and provides actionable data, but it’s just a quantification formula at the end of the day.”

The latest generation of AI could incorporate weather information, traffic patterns, near and distant competition, in-store data relative to inside sales and promotions, the impact of earlier pricing decisions on future behaviors, local events and activities—virtually anything that has a data stream and represents an intersection with customer decision making when selecting a site to fuel alongside the fuel price.

The predictive power of AI is used by SymphonyAI to help retailers optimize shelf planning, manage fresh food orders, execute promotions and review planogram compliance to boost planned and impulse buys across locations— while also understanding basket and conversion behaviors and predicting what will happen next.

Most of all, AI is best when it finds patterns that are far from intuitive. Walch, from the Project Management Institute, cited an example from Walmart of how AI assists retail sales. “When a major weather event is on the way you can expect the sales of things like flashlights, batteries, bottled water and such,” she said. However, once they started working through different data sets, they discovered people were also buying strawberry Pop-Tarts. So, it might make sense to increase sales to position and endcap with Pop-Tarts next to a section that has the other supplies.”

Let’s put all these predictive powers of AI together in a possible use case: A fuel retailer could adjust fuel prices hourly based on traffic and competitor data, use mobile app data to send a free coffee offer to a driver filling up and auto-dispatch a fuel delivery when tank levels fall below a threshold—optimized for truck routing and fuel blending.

 

AI to Improve Operations

Beyond predictive capabilities, AI can enhance operational efficiencies.

InStore.ai uses AI to monitor cashier interactions with customers. Conversations can be monitored in real time to spot areas that need to be addressed (including basic facility issues) or excellent performance that should be rewarded. For example, are the bathrooms being cleaned following customer comments? Are the associates promoting the loyalty program? Are customers complaining about the dispensers being down? Customer privacy is not a concern because the customer is not identified in the process.

The image recognition capabilities of AI also make it ideal for automating previously tedious tasks. Petrosoft uses advanced AI for image recognition to scan invoices and receipts with the ability to recognize handwritten signatures. This process is over 99% accurate and the results still have a human review to spot the occasional error.

“We are already beta testing AI cashiers and looking at applying sophisticated AI for inventory, shrinkage reduction, the forecourt and pricing,” said Munz. “Then we can combine that with economic indicators and, depending on how intuitive the owner is, we can start connecting the service vendors to find out if there are any inefficiencies in the logistics chain.”

Meanwhile, Trinium Technologies uses AI to ensure that order data is directly converted into dispatchready delivery orders that carry throughout the system, creating a reliable audit trail for invoice validation. It enhances customer service by allowing businesses to accept orders in any format, ensuring a smoother and more flexible intake process. Importantly, automation facilitates knowledge transfer by standardizing order processing, shifting “tribal knowledge” to “system knowledge.”

 

Data Security Is Critical

“If you asked me two years ago if we had any AI, I would have said 100% yes,” said John Coyle, ADD System’s vice president of sales. “We have forecasting that looks at the weather [for heating fuels] and analyzes realworld results and makes adjustments. We are already monitoring live traffic with our dispatching; we have business intelligence reporting and machine learning. But now our definitions of AI are changing constantly, and we are working to add some of the more advanced capabilities—but in a measured approach.”

A major concern during this process is ensuring data safety. AI gets smarter by analyzing all data at its disposal—and that could include your confidential information if you are not careful.

“AI needs data for its models, but our customer’s information is sacred,” Coyle said. “So how do we share that with the right AI that doesn’t share it with others that shouldn’t get it? There’s that balance of how you harness the power of collective data while keeping your information secure and private. It’s important to slow down and put in the right procedures and protocols to make sure that you have that safety, even with inhouse development and using paid AI models.”

He also cautioned managers in any business today where employees have started using readily available, often free AI tools to help automate various administrative or reporting tasks.

“There are companies that have employees that are jumping on the free version of tools like ChatGPT and saying, ‘Hey, build me this report’ and using private company data,” Coyle said. “Several years ago, one of our people found an entire customer list from a competitor online. There are tons of things that should never be on the web.”

He explained that ADD’s employees are instructed that if they use AI in their work, they have to let management know what they are doing, what they are using and how they are using it.

 

What To Consider

What considerations should a retailer or marketer keep in mind when considering potential AI-promoted solutions? Transformation and innovation specialist Sherzod Odilov provided some guidance in a recent article in Forbes: “Ask yourself, ‘Is this something that will provide a real benefit, or am I being swept up by the hype?’ Evaluate its potential ROI carefully and consider piloting small AI projects before fully integrating them into your operations.”

There are clearly top-notch vendors using quality AI tools to incorporate sophisticated AI in their solutions. However, AI is still in its “Wild West” period, in which there still are a number of vendors improperly applying AI as a marketing term more than a technology enhancement.

In the end, Odilov suggests that businesses look at the capabilities offered by a solution provider and see if they enhance operational efficiency in a way that is currently not being met by existing solutions. In other words, separate the current needs from the hype of the future. Validate claims and consider solutions based on current capabilities versus future promises.

 

Keith Reid is editor-in-chief of Fuels Market News. He can be reached at [email protected].