AI agents may be getting smarter. But in finance, intelligence still hinges on data quality.

Daloopa, which positions itself as a financial data layer for the “agentic era,” has published new research benchmarking leading AI agent systems on real‑world financial retrieval tasks. The headline finding: when powered by structured financial databases instead of public web inputs, top‑tier agents saw accuracy jump to roughly 90%—an improvement of up to 71 percentage points.

The 90% Ceiling—and What’s Missing

In controlled “FinRetrieval” tests—financial retrieval tasks requiring accurate extraction of company metrics and disclosures—agents performed dramatically better when pulling from structured, purpose‑built financial datasets.

When relying on public web inputs, accuracy lagged significantly. With structured data, performance rose to around 90%.

That’s a meaningful leap. In investment research, moving from coin‑flip reliability to near‑professional accuracy changes the risk calculus.

Fiscal Calendars: A Surprisingly Hard Problem

One revealing insight: all three agent systems performed better on US companies than on non‑US firms. Why? Fiscal year alignment. Most US public companies report on a December fiscal year, aligning with the calendar year. Many non‑US companies use non‑December year‑ends. That variance introduces complexity when agents attempt quarter‑over‑quarter or year‑over‑year comparisons.

Without standardized normalization, agents can misinterpret periods, mismap financial metrics, or surface inconsistent figures.

Data as the Real Differentiator

Daloopa’s pitch is clear: financial AI accuracy is fundamentally a data access problem.

The company delivers structured, audit‑ready financial data covering more than 5,000 public companies globally. It claims to provide up to 10 times more data points per company than other providers, with each data point hyperlinked to its original source for traceability.

“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.”

The implication: as AI agents become embedded in research desks and hedge fund workflows, structured data providers could become as strategically important as model developers.

Integrations With Frontier AI Platforms

Daloopa’s infrastructure is already being integrated into major AI ecosystems. The company recently announced a Model Context Protocol (MCP) connector with OpenAI, enabling ChatGPT workflows enriched with Daloopa’s structured financial data. It has also partnered 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.

Get in touch with our fintech experts