Daloopa Benchmark Report Shows AI Agent Accuracy in Financial Retrieval Significantly Increases When Using Structured Data

Testing 500 real-world finance questions, Daloopa found accuracy gains of up to 71 percentage points when agents used an auditable financial database instead of the public web

NEW YORK, Feb. 10, 2026 /PRNewswire/ — Daloopa, the trusted financial data layer for the agentic era, today announced the publication of its latest research report, Benchmarking AI Agents on Financial Retrieval. The study examines how well leading frontier agent systems actually perform on real-world financial research tasks when tested on 500 financial questions.

AI agents are only as smart as the data they can retrieve. Daloopa’s latest research shows how three LLM-powered agent systems—OpenAI’s Agents SDK with GPT-5.2, Anthropic’s Agent SDK with Claude Opus 4.5, and Google’s ADK with Gemini 3 Pro—saw accuracy jump to roughly 90% (up to a 71-point improvement) during financial retrieval (“FinRetrieval”) when pulling from a structured database versus public, web-sourced inputs that are often unreliable. The findings help explain why AI agents continue to struggle in high-stakes domains like finance despite advances in reasoning.

Agents that can autonomously search, reason, and retrieve data represent a significant step forward for financial research.

The findings also reveal that improving from 90% to 99%+ accuracy requires better infrastructure around models, including solving common problems related to fiscal calendars and naming conventions. For example, all three models tested performed better on US companies than non-US companies, as most US companies use December fiscal year-ends—the standard calendar alignment—while non-US companies more often have non-December year-ends.

Daloopa addresses these infrastructure challenges by delivering structured, audit-ready financial data purpose-built for AI and agentic workflows. The platform covers 5,000+ public companies globally and delivers up to 10 times more data points per company than other providers, with each datapoint hyperlinked to its original source for auditability.

“Our latest benchmark research underscores the necessity of equipping AI agents with high-quality data for FinRetrieval,” said Thomas Li, CEO of Daloopa. “Accuracy in AI-driven finance isn’t just a model problem, it’s a data access problem. Daloopa builds solutions that tackle our customers’ biggest challenges, providing the core data infrastructure for reliable AI and agentic workflows in finance.”

The leading global AI companies trust Daloopa, which has recently announced integrations with frontier AI platforms, including a Model Context Protocol (MCP) connector with OpenAI that enriches ChatGPT users’ workflows with its data. This follows on the heels of Daloopa’s similar partnership with Anthropic’s Claude for Financial Services. Daloopa’s MCP also powers analytical AI workflows ranging from hedge funds identifying quarter-over-quarter inflections and simulating scenarios, to equity researchers creating reports with full source traceability.