Anyone who’s run deliveries through regional Victoria knows the pain. It’s not like Melbourne, where you’ve got a grid, predictable traffic, and a depot every few kilometres. Out here you’ve got single-lane roads, livestock crossings, bridges with weight limits, and the odd gravel stretch that Google Maps thinks is a highway.

So when Mitchell Freight — a Bendigo-based logistics company running about 40 trucks across central and western Victoria — told me they’d started using AI to plan their routes, I was curious. And a bit sceptical, honestly. Most route optimisation software is built for urban networks. Regional is a different beast.

The problem with off-the-shelf routing

Mitchell Freight had tried two commercial routing platforms before. Both were designed for metro delivery networks. They’d suggest routes that looked efficient on paper but fell apart in practice.

“The software would route a truck through a road that’s technically open but hasn’t been graded in six months,” said operations manager Paul Grech. “Or it’d ignore the fact that you can’t get a B-double through the roundabout in Maryborough without taking out someone’s fence.”

Regional roads have constraints that don’t show up in standard mapping data. Seasonal flooding closes routes. Harvest season means slow-moving machinery on the Midland Highway. And there are customer-specific quirks — one farm requires access through a particular gate that’s only wide enough for a single-axle truck.

Building something that actually understands regional

About eight months ago, Mitchell Freight started working with one firm we talked to on a custom AI routing system. Rather than relying purely on standard map data, they fed in years of their own delivery records — actual routes their drivers had taken, with notes on road conditions, delivery times, and seasonal patterns.

The AI model combines three data sources: real-time road condition feeds from VicRoads, historical delivery data from Mitchell Freight’s own fleet management system, and weather forecasts from BOM. It recalculates optimal routes every morning at 4am, before the first truck leaves the depot.

“The first couple of weeks we ran it alongside our manual planning,” Grech said. “By week three, the AI routes were consistently better. Not by a little — by a lot.”

The numbers

Mitchell Freight shared some stats from their first six months:

• Average daily kilometres per truck dropped 14%. The AI was finding shorter practical routes that human planners missed, partly because it could process hundreds of road combinations simultaneously.
• Fuel costs down 11%. Not quite as high as the distance reduction because some of the new routes used roads with lower speed limits, meaning slightly more engine hours. But still a significant saving.
• Late deliveries fell from 8% to under 3%. The AI’s time estimates turned out to be more accurate than the planners’, largely because it accounted for seasonal slowdowns that humans forgot about.
• Driver satisfaction went up. This one surprised me. Drivers initially pushed back — nobody likes being told how to drive by a computer. But the routes were genuinely better, and drivers were spending less time on rough roads. Two drivers specifically mentioned that they appreciated not being sent down Eddington-Tarnagulla Road during winter anymore.

What this means for regional logistics

I think there’s a broader lesson here. The AI tools getting the most traction in regional areas aren’t the flashy, general-purpose ones. They’re purpose-built solutions that incorporate local knowledge.

Mitchell Freight’s system works because it was trained on their specific operating environment. It knows that the Calder Highway gets congested around Castlemaine on market days. It knows that certain customers only accept deliveries before 10am. It knows that the bridge over the Loddon at Laanecoorie has a 15-tonne limit.

That kind of contextual understanding doesn’t come from a generic SaaS product. It comes from combining AI with domain expertise.

The cost question

I asked Grech whether the investment made financial sense for a company their size. He was candid: “It wasn’t cheap to set up. Six figures. But we’re saving roughly $30,000 a month in fuel and driver overtime alone. The payback period was under five months.”

That’s a compelling case, particularly for a business where margins are tight and fuel is one of the biggest variable costs.

Not every regional logistics company can justify a custom AI build. But as these tools get cheaper and more modular, I reckon we’ll see more uptake. The regional freight sector in Victoria is worth watching.

The bigger picture for regional tech

Stories like Mitchell Freight’s matter because they show that AI isn’t just a Melbourne CBD phenomenon. Regional businesses have real operational problems that AI can solve — often more effectively than in the city, because the inefficiencies are larger and the data advantages of local knowledge are more pronounced.

If you’re running a regional operation and you’ve written off AI as something for big corporates, it might be time to reconsider. The tech has matured, the costs have come down, and the companies getting in early are building genuine competitive advantages.

I’ll be keeping an eye on Mitchell Freight. If their next move is what Grech hinted at — using the AI to predict vehicle maintenance needs based on route conditions — that could be even more interesting than the routing itself.