# MLOps and GenAIOps

> How MLOps, LLMOps, GenAIOps, platform engineering, and agent operations relate to the five BROCS operating scopes.

Source: https://brocs.fyi/resources/mlops-genaiops/
Framework: BROCS (Build, Run, Observe, Control, Secure), brocs.fyi. CC BY 4.0, attribute to brocs.fyi.

---


MLOps and GenAIOps provide practices for repeatable AI development, deployment,
evaluation, and monitoring. BROCS places those practices inside a broader enterprise
operating model that also covers access, runtime ownership, change control, cost,
security, and compliance evidence.

## Existing operations still apply

Microsoft defines generative AI operations, or GenAIOps, as an extension of existing
MLOps investments. The same guidance notes that GenAIOps is sometimes called LLMOps.[^genaiops]

Many generative AI workloads begin with a consumed foundation model. Their operating
workload includes prompts, orchestration, grounding data, indexes, evals, application
code, and sometimes agents. These artifacts need versioning, deployment, monitoring,
and rollback alongside any model configuration.

## Relation to BROCS

| Practice | Primary focus | BROCS relationship |
|---|---|---|
| MLOps | Data and model pipelines, experimentation, deployment, inference, and monitoring | Supplies practices mainly to Build, Run, Observe, and Control |
| GenAIOps or LLMOps | Prompts, orchestration, retrieval, grounding data, indexes, evals, and generative-model use | Extends Build, Run, Observe, and Control beyond the model artifact |
| Platform engineering | Shared tools, self-service paths, repeatability, and secure defaults | Supports organization-wide Build and Run capabilities |
| Agent operations | Tool access, state, traces, identity, authorization, action limits, and recovery | Requires coverage across all five BROCS scopes |

These labels overlap, but they emphasize different responsibilities. BROCS checks the
work around them and the handoffs between teams. A strong MLOps practice can coexist with
weak access control, an unsupported builder experience, missing agent traces, or
compliance evidence assembled after an incident. In BROCS, agent operations means the
runtime, identity, state, authorization, traces, action limits, and recovery required for
agents that use tools or take actions.

## Apply the crosswalk

For an existing AI workload, record the artifacts and services already governed by MLOps
or GenAIOps. Then use the [BROCS assessment](/resources/assessment/) to inspect the
organizational responsibilities around that work, including supplier and business-team
handoffs.

See [enterprise AI operations](/enterprise-ai-operations/) for the wider operating model
and [adjacent practices](/resources/adjacent-practices/) for the DORA, SRE, platform,
security, governance, and FinOps sources behind BROCS.

[^genaiops]: Microsoft, ["Generative AI operations for organizations with MLOps investments"](https://learn.microsoft.com/en-us/azure/architecture/ai-ml/guide/genaiops-for-mlops), Azure Architecture Center, updated October 17, 2025.