Owning Workflows Builds Industrial AI Moat

This is a bet that the moat in industrial AI will come from control of the work, not elegance of the software. In factories and engineering teams, the hard part is not just producing a model output, it is getting access to CAD files, simulation data, plant telemetry, approval chains, and the authority to change how work gets done. Owning assets or doing deep deployments turns messy implementation into proprietary data, embedded workflows, and direct operating gains that a cleaner SaaS model would leave with the customer.

• The product direction already reflects this. Prometheus is built to work across CAD, PLM, ERP, simulation, and procurement tools, and the General Agents acquisition added computer use capability so the system can click through old software instead of waiting for perfect APIs. That is less pure software, but much harder to rip out once it runs real programs end to end.
  1 sacra 2 aboutamazon
• The closest competitive threat is not a better demo model, it is an incumbent or vertically integrated operator with real production access. Siemens and NVIDIA already sit inside factory and lifecycle software, while Amazon is training robotics systems on its own fulfillment network at million robot scale. In this market, data and workflow control compound faster than model prestige.
  1 sacra 3 siemens 5 aboutamazon
• The cleanest comparable is Berkshire style ownership applied to industrial AI. Instead of charging a software fee and sharing upside, the company can buy or tightly control manufacturers, improve yield, throughput, and labor efficiency, then keep the full value of those improvements while using the same environments as training grounds for better models.
  1 sacra

If this works, industrial AI will look less like enterprise software and more like a new kind of operating company. The winners will be the groups that own the richest production environments, learn from them continuously, and turn those learnings into both better models and better businesses, making defensibility stronger with every deployment and acquisition.