M-OPS OMI: integrating AI into the management system you already run

The M-OPS OMI (Operations Management Integration) framework is Maestria OPS’ proprietary framework for integrating AI into the management systems that high-hazard operators already run. It is built on the requirements, recommendations and guidance of recognised AI and safety sources. Item by item, it establishes whether the existing system already controls the AI, needs modifying, or needs a new element, and it tests that the result holds where AI is involved, with particular attention to the barriers that prevent major accidents.

Engineer's desk with management system binders, a piping and instrumentation diagram, a laptop showing a chart and a checklist, with a process plant through the window

M-OPS OMI, item by item: three possible outcomes, then a test that the result holds where AI is involved.

Why not add an AI governance layer?

The usual answer, and the one generic AI tools recommend, is an AI policy and AI procedures added on top of the operating management system, which itself stays unchanged. The layer looks complete on paper. Meanwhile the management of change procedure still does not treat a model update as a change, the barrier register does not show that an AI now sits in a barrier’s decision path, and the workforce is told about the AI tool instead of being consulted before it goes in. The places where AI changes risk are left untouched.

M-OPS OMI works inside the system you already run. It is built on the ISO management-system structure and the elements high-hazard industry documents add, so it fits any operator’s system. The first step of an engagement is mapping M-OPS OMI onto your own management system; each item then gets one of the three outcomes above.

A layer is the right place only for what the existing system has no home for, such as the quality of the data fed to a model or the transparency of its outputs. Everywhere else, M-OPS OMI integrates.

What does M-OPS OMI look for in practice?

Four illustrative cases, not real incidents, show the kind of gap M-OPS OMI is designed to find:

Corlinda: a predictive maintenance model supports a deferral, and a cold box fails outside its view.
Vinmora: an AI model trained on the next field understates the risk of hole collapse.
Linmara: an AI soft sensor passes acceptance tests that covered normal conditions only.
Quillmere: a complete AI governance layer, and a management system that never changed.