# How Agents Cut Hour-Long Data Checks to 5 Minutes - Amol Kapoor & Clifford Ressel, Nori Agentic

## Executive summary

The shift of capital into private assets (e.g., land, private equity) has created significant data management challenges, as critical asset details are often trapped in unstructured formats like PDFs, emails, and call notes. Traditional Security Master (SecMaster) maintenance, which requires manual auditing of every data change, is labor-intensive and slow. Nori Agentic proposes using cloud agents—remote, specialized AI models—to automate the data investigation process. These agents can autonomously source change information, cite sources, recommend keeping or discarding data, and notify owners, drastically reducing manual back-office work from hours to minutes.

## Key takeaways

- AI Adoption is a Collaboration Problem: Successful AI adoption requires meeting in the middle between solution buyers and engineers who build foundational blocks, working closely with domain experts who can make the intuitive leap from AI concepts to specific business problems.
- Aim for 95% Automation, Not 100%: In complex fields like finance, it is more effective to keep a human in the loop (e.g., spending 5 minutes reviewing an agent's work) than attempting 100% end-to-end automation, which is often impractical.
- The Cost of Accidental Complexity: Agents are highly effective at handling 'low-complexity, high-effort' work (like standardizing data from varied formats), allowing human teams to focus on truly complex problems.

## Technical details

- Security Master (SecMaster): The SecMaster is a critical back-office master database serving as the single source of truth for all securities or instruments within an asset manager's investment universe. It is essential for risk management, P&L accounting, and regulatory compliance.
- Private Asset Data Challenges: Unlike public assets, private investments lack centralized identifiers (like ISINs or CUSIPs). Information (e.g., payment details, rights) is scattered across unstructured sources: PDFs, email attachments, Zoom call transcripts, and one-off agreements.
- Cloud Agents for Data Auditing: A cloud agent is a coding agent (e.g., Claude Code, ChatGPT) running in a remote environment. It can be configured with tools (Box for PDFs, public APIs like Bloomberg, email access) to autonomously investigate data changes in the SecMaster. It assembles an explanation, provides citations, and recommends keeping or discarding the change, notifying the owner via Slack.
- Agentic Infrastructure: The infrastructure should be generic and flexible, allowing the same cloud agent platform to power multiple functions (e.g., SecMaster auditing, audit agent, code review agent). Customization comes from natural language instructions and environment configuration, not bespoke coding.

## Practical implications

- Reduces the manual time required for data change analysis in asset management from an average of one hour per change to approximately five minutes.
- Enables asset managers to maintain clean, auditable SecMasters despite the increasing volume of data from unstructured private sources.
- Shifts the focus of back-office teams from manual data munging and searching through documents to reviewing and validating AI-generated recommendations.

## Topics

Financial Technology (FinTech), Data Engineering, Artificial Intelligence (AI), Cloud Computing, Back-Office Automation, Nori Agentic, AIE CODE summit

Source: https://www.youtube.com/watch?v=kIlVNGZZr8o
