AI That Changes What the Operation Does Next

Gravitate's VP of Marketing, Joel Davies, recently joined industry leaders at the D1 Expo for a conversation about what AI looks like when applied to real fuel operations.

Joel shared how AI can improve the high-frequency decisions operators make every day while identifying exceptions before they become operational problems. What follows is an edited version of his responses from that panel.

Start with a decision that actually got better

Everyone is talking about AI. Fewer people can point to a case where it delivered a measurable result.

A good example is load planning at Sheetz. During our proof of concept, they ran a human-versus-machine test. One of their best dispatchers took about 44 minutes to build a set of loads. The AI built those same loads in about 90 seconds.

The dispatcher still reviewed the plan and made a few adjustments. That part matters. The point was never to remove the dispatcher. It was to give an experienced person a strong starting point in a fraction of the time.

That is what measurable AI value looks like. It makes a specific decision faster or better, and you can compare the result against how the operation performs today. It is not AI for the sake of AI. It takes a complex, repetitive task and gives good people more time for customers, exceptions, and the situations that genuinely require their judgment.

The real shift is from separate decisions to connected ones

Decision support is not new in this industry. We have had forecasting, optimization, and predictive models for years. But those tools handed one person a suggestion, and that person still did all the work.

Think about what actually happens in a day. A rack price changes. An allocation gets cut at 2:00 in the afternoon. A terminal goes down. That single event affects supply, transportation, the carrier, the driver, and ultimately the customer.

Historically, each group saw its own piece of the problem and responded separately. AI can help everyone work from the same operational picture. It can identify the affected deliveries, evaluate alternate supply points, understand the cost and service implications, and give the network coordinator and the local carrier the context they need to respond.

That is especially valuable in a network model, where you have national reach, more supply and capacity options, and local operators who understand their markets. AI does not replace local expertise. It puts local expertise to work faster.

Where to place the first investment

Do not start with a broad AI initiative. Start with one decision the business makes over and over again.

There is a version of this where you go hunting for wasted time: the report that takes two hours to build, the five emails someone reads to understand what happened. That is real, and it is worth fixing. But time savings are the smaller prize. The bigger returns come from improving the decision itself.

Look for a decision with a clear financial or service impact and enough volume to measure whether you are getting better. In fuel distribution, that might be building loads, selecting a supply point, quoting a customer, or calculating margin.

Saving someone an hour is nice. Improving a decision that gets made five hundred times a day is a business case.

Start with AI making recommendations. Compare those recommendations against the current process, keep your operators involved, and expand from there. The goal does not need to be complicated: make one important decision measurably better.

Move collaboration ahead of the problem

Today, a lot of collaboration starts after someone discovers a problem.

Picture a driver heading toward a terminal that has just gone down. The carrier may not learn about it until the driver reaches the rack. Then the supplier, carrier, coordinator, and customer all start making calls.

AI can identify the affected load, evaluate another source, check the cost and allocation impact, send the diversion to the carrier, and update the customer's ETA before the truck is ever in the wrong place.

The same applies to the paperwork. Instead of someone reconciling BOLs and invoices days later, exceptions get flagged the same day. Carriers get paid faster, and nobody spends Friday chasing documents.

For a carrier or distributor, this is where being part of a network pays off. AI gets better with more data and more options, and a connected network gives you both. The load you could not cover becomes visibility into who can. The backhaul you did not know existed becomes revenue. You do not have to build any of that yourself. You have to be connected to it, with data clean enough to participate.

Five years out, capacity gets planned differently

Most companies build a plan today around the trucks and drivers they can see immediately. When demand shifts or capacity tightens, they start making calls.

Five years from now, dedicated capacity and qualified network capacity will be planned together, much earlier. AI will forecast where demand is building, where a supply or capacity constraint may develop, and which partners are positioned to help. Instead of waiting for a customer to run low or a dispatcher to discover a gap, the network can begin preparing in advance.

The routine order may begin to disappear along with it. Rather than someone watching inventory, deciding they need fuel, and placing an order, the system forecasts the need, generates the order, selects the source, and starts planning the delivery before anyone has to ask. Operators still set the rules: service levels, price tolerances, delivery windows. But the work shifts from creating and processing orders to managing a continuously replenished network.

For carriers and distributors, that world is good news if you are ready for it. Loads show up earlier with more planning time. Volume gets smoother. Runouts and emergency dispatches, some of the most expensive miles in this business, start disappearing.

Getting ready is specific: digital documents instead of paper, tank monitors instead of phone checks, and systems that can share status without someone making a call. The participants who are easiest to connect to will get the first look at that forecasted volume.

Keep humans where judgment lives

You need people wherever judgment, accountability, or relationships matter.

AI can evaluate an enormous number of options quickly. It does not always understand why a particular customer matters, why an operator might accept a higher cost today, or why a technically correct answer could damage an important relationship.

The key is not to make someone approve every AI recommendation. That just creates another inbox. Let the system handle routine decisions within clear boundaries, and bring people in when something falls outside those boundaries or the trade-offs are not obvious.

The human role does not disappear. It moves toward the work where human judgment creates the most value.

What companies are underestimating

It is easy to think of AI as another software feature. Companies are underestimating how much it will change the operating model.

Most businesses are still organized around separate departments and separate systems. But the decisions are not separate. Pricing affects supply. Supply affects transportation. Transportation affects inventory and customer service. Optimizing one department in isolation can create a problem somewhere else.

The biggest gains will not come from adding AI to each department independently. The winners will run pricing, supply, and dispatch as one connected decision, not three departments passing spreadsheets back and forth.

That holds whether you are a supplier, a distributor, or a carrier. The advantage will not automatically go to the biggest company. It will go to the most connected one, the business whose data is clean enough and whose systems are open enough to plug into the decisions happening around it.

The myth worth pushing back on

The myth I push back on is that the main value of AI is cutting headcount.

I understand the concern. But in transportation and fuel distribution, most operators are already trying to do more with limited people. The bigger opportunity is capacity and consistency: helping the same team manage more complexity, make decisions earlier, and spend less time on repetitive work.

That may allow a business to grow without adding people at the same rate. It does not mean experienced operators, dispatchers, drivers, or carrier partners disappear. Their time shifts toward exceptions, customer relationships, and the decisions where human judgment really matters.

In a network model, the value is not fewer participants. It is more productive, better-connected participants. AI should take work away from people. It should not take the operation away from them.

What to do in the next 12 months

Pick one operational decision and establish a clear baseline. Load construction. Terminal-outage response. Capacity matching. Proof of delivery through settlement. BOL and invoice reconciliation.

How long does it take today? How often is it made? How many people touch it? What happens when it is wrong? What would improvement be worth?

Then test AI against that baseline in a controlled way. Keep people involved, measure the results, and learn where the technology performs well and where it does not.

You do not need to transform the entire company in the next 12 months. You do need to start building practical experience with AI inside the operation.

Because in the end, the test is not how impressive the AI looks. It is whether the operation performs better.