Automating Investment Presentations with Generative AI & AWS

This case study demonstrates how a multi-agent AI architecture can transform labor-intensive financial reporting into a high-speed automated process while maintaining enterprise-grade security.

ClientLeader in Investment Banking

PublishedMay 2026

IndustryFinancial Services

The project addresses the critical bottleneck of manual data synthesis by leveraging a cloud-native, serverless approach to generate professional investment decks. By integrating Retrieval-Augmented Generation (RAG), the solution ensures absolute data integrity and massive operational time savings.

Our customer is a specialized investment company that focuses on high-stakes capital markets and asset management. Their core operations involve the synthesis of massive amounts of internal data, market research, and financial reports into high-quality, professional investment presentations (pitch decks) for their clients.

The company's priority is to maintain absolute data integrity and professional formatting standards while adhering to strict enterprise-grade security protocols. By implementing a secure, multi-agent generative AI architecture, we successfully reduced the time required to generate a first-draft presentation from several hours to under 2 minutes, transforming the company's operational efficiency.

Before implementing the solution, the company’s analysts spent a disproportionate amount of time on manual labor. The process of creating investment decks faced several critical bottlenecks:

Manual Data Synthesis: Analysts had to manually comb through hundreds of internal PDF/DOCX files and cross-reference them with live market data.
Time-Intensive Drafting: Creating the "first draft" of a single slide—extracting the right data points and formatting them into a corporate template—typically took several hours per analyst.
Version Control and Consistency: Maintaining brand consistency and ensuring the most recent data was reflected across all slides was a constant struggle.

To address these challenges, we designed and implemented an intelligent, automated platform built on a fully serverless, event-driven architecture on AWS. This ensures maximum scalability, security, and cost-efficiency.

To ensure precision and reliability, the architecture is divided into three functional layers:

Secure Knowledge Management (RAG) We implemented a Retrieval-Augmented Generation (RAG) pipeline to give the AI access to private internal documentation. Amazon OpenSearch acts as the vector database, while Amazon Bedrock Knowledge Bases manages the indexing of files stored in Amazon S3.
Intelligent Orchestration & Multi-Agent Reasoning A central orchestration layer powered by AWS Lambda manages the logic. We utilized Amazon Bedrock (Claude 3.5 Sonnet) for its superior reasoning capabilities and ability to output structured JSON data. To handle gaps in internal data, we integrated the Tavily AI Search API for autonomous market research.
Automated Assembly Engine The system doesn't just write text; it builds production-ready files. A dedicated Python-based engine takes the structured AI output and injects it into pre-defined corporate PowerPoint templates, automatically generating charts and tables.

The project was a resounding success, proving that Generative AI can be securely integrated into a high-compliance investment environment.

Drastic Time Reduction: The time required to generate a first-draft slide deck was reduced from hours to under 2 minutes, allowing analysts to focus on high-level strategy rather than data entry.
Cost Efficiency: By utilizing a serverless approach, the customer pays only for the compute time used during generation, eliminating the need for expensive, always-on GPU instances.
Scalability via IaC: The entire environment is managed via Terraform (Infrastructure as Code), enabling the client to deploy identical, isolated environments for different departments or security tiers in minutes.
Enterprise Security: By keeping data within the AWS perimeter and using AWS Cognito for identity management, the solution met the company’s stringent data privacy requirements.

The project provided the customer with the confidence and the technical foundation to move toward a full-scale production rollout, transforming their digital workspace into an AI-augmented powerhouse.

AWS SERVICES

Amazon Bedrock

AWS Lambda

Amazon S3

AWS Cognito

Key Takeaways

< 2 Minutes

Drafting Time

100%

Serverless Architecture

Igor is our go-to person for cloud services.

Ask him anything.