---
id: "concept-agentic-workflows"
type: "concept"
source_timestamps: ["§ Here Come the Agents"]
tags: ["ai-agents", "automation", "financial-analysis"]
related: ["entity-recon-ai", "entity-aws-bedrock-agents", "framework-agentic-report-generation"]
definition: "Hierarchical AI systems where a supervisor agent coordinates multiple sub-agents to autonomously execute complex, multi-step business processes."
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"
---
# Agentic Workflows in Finance

## Agentic Workflows in Finance

Moving beyond simple conversational **copilots**, [[entity-moodys|Moody's]] aggressively pursued **agentic workflows**, deploying its own multi-agent systems well ahead of peak industry buzz.

Its internal initiative, [[entity-recon-ai|Recon.AI]], uses a **hierarchical structure**: a **'supervisor AI agent'** manages a team of **'AI sub-workers.'** This multi-agent approach is specifically designed to generate **comprehensive financial risk reports**.

By breaking complex analytical tasks into pieces and distributing them among specialized agents, Moody's compressed a task that would traditionally take an experienced human analyst **a full week** into just **one hour** of processing time.

**Definition:** Hierarchical AI systems where a supervisor agent coordinates multiple sub-agents to autonomously execute complex, multi-step business processes.

### Connections
- The step-by-step mechanics: [[framework-agentic-report-generation]].
- Productized internally as [[entity-recon-ai]]; showcased publicly via [[entity-aws-bedrock-agents]].
- Raises the workforce question: [[question-workforce-reduction]].
- Framed by Fauber as empowerment, not replacement: [[quote-human-empowerment]].

### Enrichment note
Third-party reporting and Moody's own messaging confirm Moody's is developing AI agents and more automated analytical workflows, but the specific **'Recon.AI'** name and the exact **'one week → one hour'** productivity claim are less independently verified. Counter-perspective: multi-step autonomous systems can **compound errors** across planning, tool selection, and synthesis — experts would want validation on error rates, human-review burden, and net-risk impact before generalizing the result.


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