How Hebbia Automated 90% of Finance and Legal Work with OpenAI’s Agent SDK

Hello agent builders,

In this instalment we are covering an agent use case that is deeper and more sophisticated than most of the agents we cover. It uses the Open AI Agent SDK (previously Swarm) that allows developers to build multi-agent systems with very sophisticated orchestration. This is the type of use case that is built by a developer team and not an individual with a no-code low code tool like Agenforce or Microsoft Copilot. Instead, it is one that would be leveraged by teams looking to build mission critical agent solutions that are required to handle complex sets of workflows. It is more of a technical read but it is important to understand what solutions are emerging on the landscape.

Hebbia - A multi-agent system for finance and legal industries

In data-intensive finance and legal industries, professionals such as investment bankers, private equity analysts, and corporate lawyers spend countless hours combing through extensive market research, contracts, regulatory filings, and virtual data rooms. This manual process is incredibly time-consuming, costly, and prone to human error which can have significant adverse impacts.

To tackle these challenges, Hebbia introduced Matrix, a multi-agent AI platform built upon OpenAI's Swarm (now Agents) SDK. Leveraging sophisticated multi-agent orchestration, Hebbia has successfully automated up to 90% of complex financial and legal workflows, dramatically accelerating decision-making processes while enhancing accuracy.

Moving Beyond Traditional AI

Traditional Retrieval-Augmented Generation (RAG) methods struggle significantly when handling offline documents and private data sources. Answers are rarely explicitly stated, limiting RAG’s accuracy and effectiveness. Recognising this limitation, Hebbia harnessed OpenAI's Agents SDK to build a distributed orchestration engine capable of handling sophisticated multi-agent workflows, effectively providing an "infinite" context window.

By leveraging multiple OpenAI models in parallel—including GPT-4o, o3-mini, and o1—Hebbia’s Matrix platform achieves an impressive leap in accuracy. Benchmark tests showed Matrix boosting accuracy from 68% with traditional RAG to 92%, spanning both quantitative and qualitative deep research tasks.

Inside Hebbia's Matrix: A Multi-Agent Approach

At the heart of Hebbia’s innovation lies OpenAI’s Agents SDK, a versatile, open-source framework designed specifically for building collaborative multi-agent systems. The SDK simplifies the orchestration of multiple AI agents, each specialized for distinct roles or tasks.

Core Capabilities of the OpenAI Agents SDK:

• Agents: Customised AI “actors” configured with specific instructions, roles, and tools.

• Handoffs: Agents can delegate tasks seamlessly to specialised counterparts.

• Guardrails: Built-in safety checks and validations that ensure reliability.

• Tools Integration: Agents utilise built-in or custom-developed tools to execute complex actions such as database queries, web searches, and API calls.

• Agent Loop Automation: Automatically manages agent interactions until tasks are completed, simplifying complex workflow orchestration.

• Advanced Tracing & Debugging: Built-in visualisation tools to inspect and optimise agent performance.

Hebbia’s Matrix leverages these capabilities to create sophisticated "agent swarms," coordinating multiple agents in parallel. Tasks are intelligently delegated—complex queries are broken down, routed to the most appropriate model, processed comprehensively, and synthesised into clear answers supported by full citations.

Real-World Impact in Finance and Law

Hebbia’s agentic system has delivered significant, measurable impacts across financial and legal sectors:

• Investment Banking: Saves 30-40 hours per deal by automating marketing materials, client preparation, and counterparty responses.

• Private Credit: Automates extraction of loan terms, reducing days of manual work.

• Private Equity: Cuts 20-30 hours per deal in due diligence and research processes.

• Law Firms: Achieves 75% time savings reviewing credit agreements, saving around $2,000 per hour in legal fees.

But the value isn’t confined merely to efficiency gains. Professionals using Matrix are achieving tasks previously impossible due to human constraints, like synthesising vast amounts of historical data and referencing past deal structures dynamically during active negotiations.

Accelerating Adoption and Usage Growth

Since Matrix’s deployment, adoption has skyrocketed, underscored by impressive usage growth. For instance, in just one month (March 2025), finance and legal professionals processed more unstructured data through Hebbia’s platform than in the previous 12 months combined—exceeding 231 million pages.

The Future of Enterprise AI: Workflow Integration

Hebbia’s successful integration of the OpenAI Agents SDK signals a shift in enterprise AI. Competitive advantage no longer lies solely in raw computational power or model complexity but rather in how seamlessly AI integrates into real-world workflows. Hebbia’s Matrix showcases the power of intelligent, multi-agent orchestration for delivering nuanced, defensible insights at speed and scale.

As OpenAI’s Agents SDK evolves—expanding across different programming environments and becoming increasingly accessible—the potential for multi-agent systems like Hebbia’s Matrix to reshape industries grows exponentially.

In Hebbia’s words, “We're not just building a chatbot; we're creating an agentic operating system that tackles the world’s most complex work.” Powered by OpenAI’s cutting-edge Agents SDK, this vision is rapidly becoming reality, transforming the finance and legal landscapes for good.

OpenAI Agent SDK

For organisations or developers interested in exploring multi-agent systems, OpenAI's Agents SDK provides a powerful yet accessible starting point. With comprehensive documentation, easy integration with Python environments, and robust community support, teams can quickly prototype and deploy sophisticated AI workflows. Whether your goal is streamlining complex tasks, enhancing decision-making, or driving innovation, the Agents SDK offers the flexibility and power necessary to unlock the full potential of agent-driven automation.