Fortescue: Scaling AI Across Autonomous Mining Operations – From Pilot to Production

Brief

We embedded with Fortescue’s AI team and delivered a suite of production-ready AI systems across a six-month pilot, then continued the engagement to bring those systems fully into production. The work spans AI Agents and Computer Vision applications running inside the fleet control environment: automating time-consuming manual workflows, flagging truck and road issues before they escalate, and surfacing optimisation opportunities the team previously couldn’t see. One automation alone is saving approximately three hours per day, and the partnership remains ongoing.

Great experience working with AIA. Speed, execution and results!

Dieter Haage, AI Product Manager, Fortescue

Problem

Running an autonomous haulage fleet at Fortescue’s scale means an enormous volume of daily data: truck performance, road conditions, fuel consumption, shift reporting, and fleet positioning. The team responsible for getting the most out of that fleet was capable, but there simply weren’t enough hours in the day. A significant portion of each shift went to repetitive, manual tasks that had to happen every day without fail, and when people do the same thing hundreds of times, errors creep in. At fleet scale, small errors have real consequences. Time spent on data handling and reporting was time not spent on optimisation, and the team was constantly reacting to issues rather than getting ahead of them. What Fortescue needed were AI systems that could take the repetitive work off their people entirely, flag problems early enough to prevent downtime, find efficiency gains that weren’t visible in the raw data, and do all of this reliably in a live operational environment.

Solution

We embedded directly with Fortescue’s AI team, working hands-on each week to design, build, and validate a suite of AI systems for their autonomous fleet control environment. Rather than building in isolation and handing over, we worked in tight cycles: prototyping fast, testing against real operational conditions, and iterating until each system was production-ready.

• Built for the operational environment, not a demo: every system was validated against real fleet data and real workflows before going live
• Errors eliminated at the source: the daily tasks most prone to human error are now handled by AI, with one automation saving approximately three hours per day at a near-zero error rate
• Proactive, not reactive: AI systems now flag issues early, giving the team time to act before problems compound into downtime or cost
• Optimisation made visible: efficiency opportunities buried in fleet data are surfaced automatically, turning invisible inefficiencies into actionable decisions
• Capability that stays: technical workshops ran alongside delivery so Fortescue’s AI team understands what was built and can manage, adapt, and extend it independently

Result

What began as a six-month pilot has grown into an ongoing partnership. After completing the initial phase, we continued with Fortescue to bring systems into full production and expand the scope of AI across the organisation, and that work is still active. On the ground, the impact is already measurable: one automation in production is saving approximately three hours per day for the fleet control team, time now available for analysis, optimisation, and decisions that require human judgement, and the same system has effectively eliminated errors on tasks that previously relied on manual data handling. At the fleet level, earlier detection of truck and road issues gives the team a genuine window to act before problems become downtime, and fuel optimisation insights that weren’t previously visible are now surfaced routinely. For Fortescue’s AI function, the engagement has accelerated internal capability, with production-grade systems running, a team equipped to manage and extend them, and a proven model for moving from idea to deployed AI at pace.

Under the Hood