Back to platform

AI Assistant

Ask a question in plain language (English, Spanish, French, German and more). The answer is built from your own portfolio data.

The AI Assistant answers the questions that no single report answers. You type a question in plain language, in English or another major language. The platform sends it to a large language model together with the relevant part of your calculated data — the same positions, transactions, prices, returns, risk and compliance results that feed every report — so the answer is built from your numbers, not from general knowledge. Use it when a question cuts across several reports, or when you want a number explained.

Example questions

5 example questions in this category. Each answer works from the same calculated data that feeds the reports.

Questions that cut across reports

Which positions lost money this quarter but grew as a share of the portfolio?

Combines figures that live in different reports (here, returns and weights) in one answer, without exporting both reports to a spreadsheet and matching them by hand.

Explaining a number

Why did the portfolio trail its benchmark last month?

Works from the performance and attribution results already calculated for the portfolio, so every figure the answer mentions can be checked in the matching report.

Your book, read together

Which holdings are closest to a compliance limit, and how large are they in the risk numbers?

Reads the compliance and risk results side by side. The assistant only reads: it does not place trades or change any data.

Asked in Spanish

¿Qué posiciones aportaron más a la rentabilidad de la cartera en lo que va de año, y cuánto pesan hoy?

Which positions contributed most to the portfolio’s return this year, and how much do they weigh today? Asked in Spanish, answered in Spanish, from the same return and position results as the English reports.

Asked in Chinese

本月哪些持仓对组合风险的贡献最大?它们离合规限额还有多远?

Which holdings contributed most to portfolio risk this month, and how far are they from their compliance limits? Asked in Chinese, answered in Chinese, from the same risk and compliance results.

Languages

Questions and answers are not limited to English. The language model behind the assistant reads and writes most world languages; accuracy is close to English in widely spoken languages and lower in languages with less written material.

Published by Anthropic for Claude Sonnet 4.5: scores on the MMLU knowledge test, translated into each language by professional translators, as a percentage of the English score (with extended thinking). Source: Anthropic, Multilingual support.
Language% of English
English100.0%
Spanish98.2%
Italian97.9%
Portuguese (Brazil)97.8%
French97.5%
Indonesian97.3%
Arabic97.2%
German97.0%
Chinese (Simplified)96.9%
Japanese96.8%
Korean96.7%
Hindi96.7%
Bengali95.4%
Swahili91.1%
Yoruba79.7%