Alibaba Cloud Releases AI Investment Research Workbench Qovest, Targeting Professional Institutional Investors

New crowdsourced testing promotional copy New crowdsourced testing promotional copy end Add App-exclusive traffic redirect

[TechWeb] November 2 news: Alibaba Cloud has officially released Qovest, an AI investment research workbench for professional institutional investors. Users submit a research request in a single sentence, and Qovest automatically queries data, collects materials, performs in-depth analysis, generates a report, and continuously tracks the research subject at a set frequency. Unlike one-off Q&A, Qovest can archive each research result in an investment research space, allowing a research project to continue onward without starting from scratch every time. Soon, Qovest will launch an enterprise edition to serve the investment research teams of financial institutions and enterprises.

Professional investors' requirements for Agents are far higher than those for everyday chat: data must be real-time and accurate, output results must have clear sources, and research conclusions must be professional and reliable. To this end, the Qovest team purchased authoritative financial databases, deployed them on Alibaba Cloud, and independently developed an Agent engineering framework (Harness) for financial scenarios, connecting the complete process from data invocation to analysis output to make delivered results more accurate and professional. In internal horizontal evaluations on the open-source financial benchmark sets FinFirst and IRAB, Qovest led mainstream products participating in the evaluation in answer quality, response speed, and cost.

"We hope to hand the time-consuming and highly repetitive work in the investment research process to AI, and give time back to the places that truly require professional judgment," said the Qovest product lead.

The Qovest homepage provides multiple investment research case templates. Users can select their own targets as research subjects, set scheduled tasks, connect their own trading notes, and obtain visualized in-depth research results. For example, users can ask Qovest to break down the per-unit value and domestic content level of each segment in the humanoid robot industry chain, obtaining an interactive industry chain map; they can also ask Qovest to estimate the investment scale of data center power facilities—adjust any one underlying assumption, and Qovest automatically recalculates.

Beyond in-depth research, Qovest can also keep research continuously tracking the market. One fund manager set up a tracking study for a certain "gold ETF" in Qovest: after defining variables such as international gold prices, the U.S. dollar index, global ETF holdings, and central bank gold purchases, Qovest automatically updates the data every week, breaks down the respective shares of exchange-rate factors and rate-hike expectations in that week's gold price moves, and generates a dynamic tracking dashboard. From then on, every time the research page for this ETF is opened, Qovest can continue the analysis from where it left off.

This experience of "not having to start from scratch every time" comes from Qovest's "investment research space." It is integrated with Qovest conversations and can automatically archive research results, personal trading notes, and questions to be verified, making research traceable and comparable. In the future enterprise edition, the "investment research space" will become a research scheduling center for collaboration and information sharing among members of an organization.

At present, the Qovest official website has opened invitation-based testing to individual users. After registering, new users receive corresponding credits for a free trial, and can subsequently choose a monthly subscription plan of 99 yuan or 199 yuan, each including a corresponding usage quota. Statement: Exclusive article from Sina.com; unauthorized reproduction is prohibited. tech_web_article_v2015_block_bottom Non-targeted 300*250 button 17/09 wenjing begin

Non-targeted 300*250 button end