By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
Fuel is usually the largest or second largest operating cost a trucking company has, and unlike insurance or equipment payments it changes every day. That makes it a natural target for software that promises savings. Some of those promises hold up. Others describe savings that were never available in the first place.
As an AI transportation consultant, the first thing I look at is whether a fleet can already answer a basic question: what did each truck burn last week, where was the fuel bought, and how many miles did it cover? If that answer takes a day of spreadsheet work, the opportunity is not artificial intelligence yet. It is getting fuel card, telematics and dispatch data into one place. Everything below depends on that.
The four places fuel money goes
Fuel spend comes down to four things: how many miles you run, how many gallons each mile takes, what you pay per gallon, and how many gallons are paid for but never reach the engine. AI tools help with all four, though not equally.
1. Buying fuel in the right place
Fuel purchase optimization is the oldest and most reliable source of savings, and it is where machine-driven planning earns its keep. The idea is simple. Along any route there are many places to buy fuel, at different prices, and your fleet has different negotiated discounts at different chains. The tool recommends where to stop and how many gallons to buy at each stop.
Two details make this harder than it sounds:
- The pump price is not the real price. Interstate carriers settle fuel tax by jurisdiction based on the miles driven in each one, not on where the fuel was bought. A cheap pump price in a low-tax state can cost more after the quarterly tax reconciliation than a higher pump price next door. Good optimization compares price net of tax.
- Your discount matters more than the posted price. The plan has to use your actual negotiated pricing, updated daily, or the recommendations are wrong.
The savings here are real, but only if drivers follow the plan. Measure compliance by truck. A fuel plan that half the fleet ignores produces half the result, and usually the reason is practical: the recommended stop has no parking, bad access or a long line. Ask drivers why before assuming they do not care.
2. Running fewer wasted miles
The gallon you never burn is the cheapest one. Empty miles and out-of-route miles are fuel spent with no revenue behind it. Optimization tools that sequence stops, respect appointment windows and suggest reloads close to the delivery point reduce both. I covered the mechanics in AI route optimization explained, and the cost side in what deadhead is in trucking.
A related use is out-of-route detection. Software compares the miles a truck actually drove against the practical route for the load and flags the gap. Some variance is legitimate: a closed road, a detour for a safe parking spot, a scale. A repeated pattern on the same lane or with the same truck deserves a conversation.
3. Burning fewer gallons per mile
Idle time
Idling is the easiest waste to see and one of the hardest to fix, because much of it has a reason. A driver waiting hours at a dock in summer heat is not misbehaving. Useful tools separate idle time by cause and location: idling at a customer during a long wait, idling during a rest break, and idling for no clear reason. The first is a detention problem to raise with the customer. The second is an equipment question about auxiliary power. Only the third is a coaching matter.
Speed and driving style
Fuel economy falls as highway speed rises, and hard acceleration and late braking cost fuel as well. Telematics data scored by a model can show each driver how they compare to peers running similar loads on similar terrain. The comparison has to be fair. A driver pulling heavy loads through hills will never match one running light freight on flat ground, and a scorecard that ignores this loses the drivers’ trust immediately.
Equipment condition
Underinflated tires, misalignment, a dragging brake or a failing injector all show up as a slow decline in fuel economy before they show up as a breakdown. Anomaly detection is well suited to this: compare each truck to its own history and to similar units, and flag the ones drifting down. This overlaps with predictive maintenance, and a good program treats the two as one data stream.
4. Stopping fuel that never reaches the tank
Fuel theft and card misuse are uncomfortable subjects, but they are a real line item. Pattern detection is good at catching the common forms:
- A purchase larger than the truck’s tank capacity
- Gallons bought that exceed what the tank could have held given the level reported by the truck
- A fuel transaction at a location the truck’s GPS says it never visited
- Two fills close together with few miles between them
- Purchases of non-fuel items or the wrong fuel type on a truck card
- A card used after hours or while the truck was parked at the yard
Rules can catch the obvious cases. Machine learning adds the ability to notice a truck or card whose behavior has shifted compared with its own baseline. Either way, treat a flag as a question and not an accusation. Sensor errors, a shared card during a breakdown or a mistyped odometer reading explain many alerts. Card-level controls such as gallon limits, product restrictions and prompts for unit number and odometer prevent more loss than any detection model.
Where the promised savings disappear
- Double counting. A vendor quotes separate savings for routing, idle reduction and driver coaching, and the sum exceeds what is physically possible. Savings overlap. Ask for a combined estimate based on your own data.
- Bad baselines. Fuel economy changes with season, freight mix, load weight and lane. Comparing a mild month to a harsh one proves nothing. Compare the same lanes and similar loads across like periods.
- Dirty data. Missing odometer entries, cards not tied to a unit and trucks swapped between drivers make per-truck numbers unreliable.
- Alerts nobody owns. If no one is assigned to review exceptions each week, the tool becomes a report that gets filed.
Recovering fuel cost is a separate job
Using less fuel and getting paid for fuel are different problems. A surcharge schedule that lags the market, or a contract with no surcharge at all, can erase operational gains. Review how your customer agreements handle it, using how fuel surcharges work in trucking as a reference, and feed your real fuel cost into your cost per mile calculation so pricing reflects it.
A sensible order of work
- Connect fuel card, telematics and dispatch data so every gallon is tied to a truck, a driver and a load.
- Tighten card controls.
- Turn on purchase optimization and track driver compliance.
- Add exception detection for theft and for trucks losing efficiency.
- Start driver feedback last, once the data is trusted, and frame it as coaching.
Fleets that follow this order tend to find that the early steps, which involve little or no AI, deliver a large share of the result. The models add value after the basics are in place.
If this kind of practical breakdown is useful, subscribe to the Freight Guru Podcast for more.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.