Mixed reality industrial training has finally moved past the pilot phase in Australian manufacturing and resources. The big change in the last twelve months is that several Tier-1 operators have rolled MR training out at scale rather than running yet another proof of concept. The picture is honest rather than triumphal.

What’s working in production deployments: training for confined-space entry, electrical isolation procedures, high-voltage equipment safety, and specific maintenance procedures on expensive rotating equipment. The common thread is high-consequence, infrequent procedures where MR provides realistic exposure that can’t safely be done on the live equipment. The ROI math works on these specific use cases. Reduced incident rates, faster competency progression, and reduced shutdown time for hands-on training are showing up in actual operational data, not just vendor case studies.

What’s not yet working at scale: general onboarding training, soft-skill training, and any training where the underlying competency has more to do with judgment than procedure. The MR experience for these is technically possible but it’s not better than well-designed traditional training. The cost-benefit math doesn’t justify the headset deployment overhead.

The headset hardware story has improved meaningfully. Quest 3 derivatives, the latest Pico industrial offerings, and the higher-end XR offerings from the Apple side of the market are all credible enterprise platforms in 2026. The hardware fragmentation has reduced enough that procurement teams can standardise on a small number of platforms rather than running compatibility matrices.

The content gap is still real. Building MR training content remains expensive and slow compared to building traditional training. The companies that have scaled MR training have generally invested in internal MR content teams rather than relying on per-procedure vendor builds. That investment is significant. Internal MR teams of 6-12 people, with serious learning design, MR development, and SME-engagement capability, are what’s actually behind most successful enterprise rollouts.

The other quiet story is the integration with Learning Management Systems. The 2024-25 generation of MR training platforms didn’t integrate cleanly with corporate LMS systems. The 2026 generation is much better. Tracking, attestation, competency progression, and audit trails work end-to-end. That sounds boring, but it’s the difference between MR training being a curiosity and being a real corporate training capability.

What’s coming: AI-generated MR scenarios. Several vendors are pushing this, and the early demos are striking, but the production-grade implementations are still early. The honest read in May 2026 is that AI-generated MR is going to be transformative on a 2-3 year horizon and is mostly demo-ware right now. The procedural correctness, the safety-critical accuracy, and the SME-validation processes aren’t yet at the standard that high-stakes industrial training requires.

For Australian operators evaluating MR training investment in 2026, the practical advice is to be specific about the use cases that actually justify the deployment, invest in internal MR content capability if you’re going to scale it, and don’t expect MR to replace traditional training. It’s a complement to it, particularly valuable for the high-consequence-low-frequency category, and that’s where the spend should focus.

The vendors that have positioned themselves around specific industrial use cases are doing better than the generalist MR training platforms. Specialisation pays in this market. The vendors trying to be all things to all enterprises are not winning the deployments.