LLM Data Labelling and Annotation

I need about 500 - 1000 rows of LLM response data consisting of context, queries and auto-generated labels verified by a 2 step human review process. The first step will involve 2 agents looking at the exact same row and a 3rd agent will be used for quality assurance. The agent must have familiarity with english language and basic general knowledge. For items containing technical content or financial content, agents must do best effort labelling and relying on the expert agent to do the work. At least 1 out of the 3 agents must be an expert in assessing technical and financial content.