---
id: "framework-agentic-report-generation"
type: "framework"
source_timestamps: ["§ Here Come the Agents"]
tags: ["ai-agents", "workflow-automation"]
related: ["concept-agentic-workflows", "entity-recon-ai"]
steps: ["Deploy a Supervisor AI Agent tasked with overseeing comprehensive financial risk report generation.", "The Supervisor Agent delegates specific analytical sub-tasks to a team of specialized AI sub-workers.", "The sub-workers process proprietary data and domain-specific analysis in parallel.", "\\\"The Supervisor Agent synthesizes the sub-workers' outputs into a final", "comprehensive report", "reducing a 1-week human task to 1 hour.\\\""]
sources: ["execution"]
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-cl-93-legacy-financial-all-in-genai"
sourceUrl: "https://hbr.org/2025/03/how-a-legacy-financial-institution-went-all-in-on-gen-ai"
sourceTitle: "How a Legacy Financial Institution Went All In on Gen AI"
---
# Multi-Agent Financial Report Generation

## Framework: Multi-Agent Financial Report Generation

Moody's approach to automating complex, week-long analyst tasks using a **hierarchical agentic architecture** (productized internally as [[entity-recon-ai|Recon.AI]]).

1. **Deploy a Supervisor AI Agent** to oversee generation of a comprehensive financial risk report.
2. The **Supervisor delegates** specific analytical sub-tasks to a team of specialized **AI sub-workers**.
3. The **sub-workers process** proprietary data and domain-specific analysis **in parallel**.
4. The **Supervisor synthesizes** the sub-workers' outputs into a final, comprehensive report — reducing a **1-week human task to 1 hour**.

### Connections
- The concept: [[concept-agentic-workflows]].
- Showcased publicly via [[entity-aws-bedrock-agents]] (Dec 2024 stage-share).
- Workforce implications: [[question-workforce-reduction]].

### Enrichment note
Consistent with broader **supervisor-agent orchestration / tool-use planning** literature, though operational details remain vendor- and implementation-specific. Caution: multi-step autonomous systems can **compound errors** across planning, tool selection, and synthesis; the '1 week → 1 hour' gain is plausible but warrants validation of error rates and human-review burden.


## Related across articles
- [[concept-autonomous-agentic-operations]]
- [[action-manage-ai-agents]]