AI-Based Fuel Intelligence System Development

I need to build a Fuel Intelligence System around the fleet.
The system should collect:
1. Truck
GPS
Odometer
Engine hours
CAN-bus data
Fuel level
Fuel consumption
Idling
2. Driver
Driver ID
Driving behaviour
Speed
Harsh acceleration/braking
Excessive idling
Route deviations
3. Fuel
Every fuel-card transaction
Litres
Rand value
Station
Date/time
Truck
Driver
GPS position
4. AI layer
The AI then calculates the expected fuel consumption for every truck and journey.
For example:
Truck normally achieves 2.4 km/L
Today's trip = 720 km
Expected fuel = 300 L
Actual fuel purchased = 390 L
Variance = 90 L
The system then investigates why.
It could determine that the additional consumption was caused by:
excessive idling
speeding
heavy traffic
route deviation
vehicle mechanical problem
abnormal fuel transaction
inaccurate fuel sensor
possible fuel theft/diversion
This is where AI becomes really valuable.
Instead of your fleet manager looking through 300 trucks × hundreds of fuel transactions, the AI tells them:
7 trucks require investigation today
18 trucks have abnormal fuel consumption
275 trucks operating within expected parameters
That is the kind of system I'd target.
One important point
Because I don't operate your own diesel tank, I'd put fuel-card integration at the very top of your requirements. A system that only measures the truck's fuel level won't give you the complete picture.
For your fleet size, I'd therefore have Geotab and MiX by Powerfleet demonstrate their systems head-to-head, with Netstar ProFleet as the South African alternative.