A team reuses the same prompts, guardrails and retrieval across projects. What is the benefit of standard components?
Why D is the answer
Consistent implementation across deployments
Standard components give consistent behavior across deployments, so teams do not rebuild the same parts.
In detail
- Standard components give consistent implementation.
- Shared components need an owner.
Use case
An illustrative scenario, not a real incident.
Four projects reuse one set of retrieval and guardrail components, so a fix reaches all of them.
See it in different ways
sequenceDiagram participant S as Service participant L as Abstraction layer participant C as Configuration participant M as Model S->>L: Request L->>C: Which model? C-->>L: Model ID L->>M: Call M-->>S: Response
mindmap
root(("Standard components"))
n1["Shared"]
n2["Prompts"]
n3["Guardrails"]
n4["Retrieval"]Key point: Standard components give consistency.
Why the other options are wrong
- A. Each team can use a completely different approach for every project
- That is the opposite of standardizing.
- B. Models no longer need to be evaluated before they are released
- You still evaluate.
- C. The monthly cost of inference falls to zero after the first project
- Inference still costs money.