Bloomberg introduced BQuant Assistant on 28 October 2026, a conversational AI coding assistant built into BQuant Desktop and BQuant Enterprise. A Terminal subscriber types a request in plain language and gets Python, Bloomberg Query Language (BQL) queries or chart code back, written inside the same environment where the analysis will run.
What BQuant Assistant does
BQuant is Bloomberg’s Python-based data science suite for financial markets. Users build investment strategies, dashboards and quantitative analyses in it, and can publish the results as interactive BQuant Published Applications that permissioned colleagues at their firm can open. BQuant Assistant sits inside that development environment. According to Bloomberg, it answers questions about BQuant, generates Python code, helps users query Bloomberg’s datasets with BQL, debugs and iterates on their work, and writes code for data visualizations.
Bloomberg describes the assistant as domain-grounded. It searches the BQuant Help Center and draws on example notebooks, libraries and how-to guides when it answers, and it can see the context of the user’s own project, workflows and code. Bloomberg says that grounding gives more relevant support for financial data analysis than a general-purpose coding assistant provides.
Advertisement
The release lists what users can ask it to do:
- Generate BQL and PyBQL queries that retrieve and analyze data across equities, funds, fixed income, FX, commodities, indices and derivatives, drawing on more than 22,000 fields that include reference data, pricing, returns, fundamentals, valuations, risk metrics, estimates and economic data.
- Write code for screens and universes across equities, fixed income, foreign exchange, commodities, indices and exchange-traded funds.
- Insert, run, debug and iterate on PyBQL or Python code in an active notebook, subject to the user’s approval.
- Answer questions about Bloomberg libraries, BQuant workflows and client-facing product documentation.
- Explain financial concepts and generate code toward models or techniques suited to a workflow.
- Recommend visualizations and write the code for them with open source libraries such as Plotly and Jupyter Widgets.
- Explain code the user selects, which Bloomberg says helps users build their Python knowledge as they work.
Approval comes before it touches a notebook
The assistant can do more than suggest code. Subject to the user’s approval, it can insert code and run it directly in the active project inside the secure BQuant sandbox. Bloomberg says users keep full control because the assistant identifies the action it proposes and asks for approval before it reads from or writes to a notebook.
Anyone who signs off on AI tools in a regulated firm will look at that step first. A coding assistant that executes code against licensed market data carries more operational risk than one that drafts text, and the release puts a person’s confirmation of each read and write inside the product itself.
Bloomberg’s chief technology officer, Shawn Edwards, framed the problem the assistant addresses: “Financial professionals are working with increasingly large and complex datasets, but turning an investment idea into a repeatable analysis can still require significant time and specialized coding knowledge.” He added that the assistant is aimed at everyone who uses BQuant, “from a portfolio manager sketching out a first query to a quant engineering a multi-asset signal.”
Who gets it, and when
BQuant Assistant will be generally available in BQuant Enterprise and in limited beta in BQuant Desktop at the end of October 2026. The two products differ in how a firm buys and runs them. BQuant Desktop is included with a Bloomberg Anywhere subscription and gives enabled Bloomberg Terminal users programmatic access to Bloomberg data and analytics in a secure sandbox. BQuant Enterprise is a fully managed, cloud-based analytics platform with large-scale compute, internal data integration and collaborative development options. Both versions will include the assistant.
The release does not give pricing for the assistant or say whether it is charged separately from the existing subscriptions.
One of several AI layers on the same data
Bloomberg places the launch in a series. The release points to ASKB, the Bloomberg Terminal’s conversational AI interface, and to Enterprise Model Context Protocol (MCP), an AI access layer for Data License Plus, Bloomberg’s next-generation Data License offering. Each reaches the same body of data by a different route: a chat window in the Terminal, a Python notebook, and a connection for AI agents that a client builds itself.
Data vendors putting AI on top of their own data is a pattern we have followed on this site. We reported on S&P Global putting credit ratings inside AI models, and on how bank AI agents are being sold on old cores, with boundaries set at runtime. In our reading, the BQuant launch fits that pattern, because the vendor that licenses the data also supplies the assistant that writes code against it.
What it means for the finance leader
For a head of quantitative research, data engineering or technology at an asset manager, bank or broker, the questions are practical, and the release answers some of them and leaves others open.
The release answers where the code runs: in the BQuant sandbox. It answers who approves actions: the user, before any read or write to a notebook. It does not say what the assistant records about prompts and approvals, how that record could be reviewed by a compliance team, or how the assistant treats a firm’s own internal data in BQuant Enterprise. Those are questions to put to Bloomberg before a rollout, in writing.
A second consideration is review. Bloomberg says the assistant can explain code a user selects, so more people will be writing and running Python, and review standards for AI-written queries become a team decision. A firm can require that any query feeding a published application is read by a second person, and can say so in its model risk policy before the beta reaches the desk.
Third, the end-of-October timing is close. Teams that already run BQuant Desktop can ask their Bloomberg representative whether they are in the limited beta, and teams on BQuant Enterprise can plan a pilot on a non-production project first, with a short list of tasks to compare: time to a working BQL query, errors caught in review, and the number of approval prompts a user accepts without reading them.
