Metacognition AI raises $10M for voice-trained robot OS, targets industrial ops
Adelaide-based Metacognition AI closed $10 million from CSIRO-backed Main Sequence to build what it calls "Windows for robots": a voice-controlled operating system that lets industrial workers train robots by talking to them.
The company launched this year. Cofounders are Anton van den Hengel (chief scientist at the Australian Institute for Machine Learning), Stephen Gould, and Paul Dalby. Current headcount: eight. They are hiring AI researchers.
What they are building
Metacognition sits underneath large language models and controls how AI agents interact with machinery. The pitch: operators can direct an open-cut mine or manage an electricity grid by speaking instructions. The system remembers commands and improves performance over time.
Initial target market is industrial systems: mining, energy, heavy machinery. The team has built simulations for these use cases but has not disclosed live deployments or customer names.
Market context
This is a crowded category. Hardware-agnostic robot operating systems and autonomy layers are attracting capital: OpenMind, Stateful Robotics, Generalist AI, Apptronik, and Figure are all raising or shipping in adjacent spaces. The $10 million round is sizable for seed-stage robotics software in ANZ but modest by global standards.
Sales reality check
No public information on enterprise sales structure, CRO, or VP Sales. No confirmed ANZ customer base. No disclosed revenue or ARR. The company is likely in early commercialization: they have funding and a pitch, but it is unclear if they have repeatable sales motion yet.
For sales professionals watching this space: robotics software requires long enterprise cycles, technical pilots, and deep integration work. If Metacognition is hiring account execs or solutions engineers in the next 12 months, that will signal traction. Until then, this is a team of eight researchers with a demo and $10 million to prove product-market fit.
Worth noting: Main Sequence has a strong track record backing deep tech in ANZ (Q-CTRL, Morse Micro, Harrison.ai). That backing matters when selling into conservative industrial buyers who want proof the vendor will still exist in three years.