From Smart Responses to Autonomous Trades
Just three years ago, LTX — a division of technology giant Broadridge Financial Solutions — launched the BondGPT tool. It was a generative AI application for the corporate bond market, which helped traders quickly find information and discover investment opportunities. It functioned as a smart assistant: answering queries, analyzing data, offering insights.
In June 2026 at the FILS USA 2026 conference, however, LTX announced a significant shift. BondGPT gained agentic capabilities — and transformed from a passive advisor into an active trading assistant, managing the entire pre-trade process: from real-time market condition monitoring through opportunity identification to trade order creation and its automatic execution.
“When we launched BondGPT, our goal was to make it easier for traders to discover information and opportunities. Agentic capabilities represent the next step — they allow for delegating tasks and smoothly transitioning from analysis to execution,” explained Jim Kwiatkowski, CEO of LTX.
What Agentic AI in BondGPT Specifically Does
According to Katie Savignano, Director of Relationship Management at LTX, and Diana Demianczuk, Relationship Manager, who discussed the new development in an interview for Trader TV, AI at LTX is moving from the generative phase (creating analyses and recommendations) to the agentic phase (acting independently).
In practice, this means a trader can give the agent a task such as: “Look for ten-year corporate bonds with a BBB+ rating, trading at a spread over 100 basis points, and where dealers are selling more significantly than yesterday.” The agent then continuously monitors the market, evaluates signals from price and liquidity data, and when a match is found, automatically fills out the trade ticket, selects suitable counterparties, and — if the trader has set up automatic execution — also executes the trade.
Crucially, all agent actions take place within clearly defined boundaries: trade size limits, human approval for key decisions, and full auditability of every step. LTX uses its own patented “show your work” approach, which ensures transparency — every agent's decision is traceable and explainable. Broadridge obtained the US patent for the BondGPT mechanism in 2025.
Who is Behind the Platform and Who Uses It
LTX is not a garage startup. It's a subsidiary of Broadridge — a company with over $6 billion in annual revenue, which provides technology infrastructure for a large part of the global financial sector. The LTX platform focuses on electronic corporate bond trading, a market that has long been the domain of phone calls and manual processes.
In early 2026, five major banks joined the platform: Goldman Sachs, JP Morgan, Morgan Stanley, Bank of America, and TD Securities. These institutions act as liquidity providers on LTX — i.e., as counterparties with whom buy-side traders (investment funds, asset managers) can trade via the platform. LTX also won the European Markets Choice Award in the Best in AI category.
Competitive Advantage Is No Longer in the Technology Itself
One of the key ideas that emerged from the interview: agentic AI is becoming a commodity. As technology becomes cheaper and widely available, the mere ownership of an AI agent ceases to be a competitive advantage. “The real advantage shifts from the technology to how it is applied,” Demianczuk stated.
Decisive will be proprietary know-how, proprietary data, and institutional knowledge — i.e., execution history, portfolio context, counterparty relationships, and internal research. These are things that competitors cannot easily replicate. In other words: an AI agent will soon be as common as a Bloomberg terminal. What will make the difference is how cleverly a company integrates it into its specific workflows.
What This Means for Europe and the Czech Republic
Although LTX and BondGPT primarily operate in the American corporate bond market, the trend of agentic AI in trading also has a direct impact on European financial institutions. European banks and investment funds face the same pressure for higher productivity without increasing headcount. Agentic AI promises exactly that: to be a "force multiplier" for the trading department — taking over repetitive routine tasks and freeing up traders for strategic decisions.
For Czech financial institutions — from large banks like ČSOB or Komerční banka to investment funds — agentic AI in trading currently represents a more distant horizon. There are several reasons: the regulatory environment (the EU AI Act introduces strict requirements for high-risk AI systems, which may include agentic trading), smaller trading volumes in the local market, and a more conservative approach to automation. Nevertheless, the direction is clear — just like algorithmic trading ten years ago, agentic AI will gradually become the standard.
The Czech National Bank (ČNB) has, after all, shown that Czech institutions do not want to lag behind — in 2026, it launched its own AI center with Nvidia chips and uses models from Mistral, OpenAI, and Alibaba for financial market supervision. The transition from supervision to active trading with AI assistance is a logical next step that will follow once proven technologies mature.
Where Further Development Is Headed
According to Katie Savignano, the next phase of agentic AI in trading is the transformation of tools like BondGPT into true trading assistants — systems that take over delegated work, continuously monitor the market, and advance tasks without constant manual input. “In an environment where most companies face pressure to do more without increasing headcount, AI is not just a productivity tool — it’s a force multiplier for the entire department,” she added.
What's most interesting about this shift: it's not about AI replacing human judgment. It's about changing the definition of a trader's job. Instead of spending hours sifting through tickers and filling out tickets, they focus on what only humans can do: strategic decision-making, relationship building, and assessing risks that fall outside algorithmic models.
Is agentic AI in trading safe? Can't it cause a flash crash?
Platforms like BondGPT are built on the principle of "controlled autonomy" — agents operate within predefined limits (trade size, price range, approved counterparties) and key decisions are always approved by a human. Moreover, every agent action is fully auditable. The risk of an uncontrolled chain of automatic trades leading to a flash crash is thus significantly reduced — unlike fully autonomous high-frequency trading systems where decisions occur in microseconds without human intervention.
Can a regular retail investor use a similar tool?
Currently, no. BondGPT is designed for institutional clients — banks, investment funds, and asset managers who trade in the corporate bond market, which is a market with minimum volumes in the millions of dollars. However, simpler forms of AI assistance exist for retail investors — for example, Robinhood launched agentic trading for retail investors in 2026, which operates on a similar principle, albeit with significantly more limited options.
How does agentic AI differ from algorithmic trading that banks have been using for years?
Algorithmic trading (algo-trading) is governed by fixed, programmed rules — "if price drops below X, buy Y." Agentic AI, in contrast, understands goals in natural language, evaluates context independently, seeks the optimal path to the goal, and can explain its decisions. For example, it can understand the instruction "find bonds that are undervalued compared to comparable issuers and prepare a trade" — something a classic algorithm cannot do because it requires an understanding of the market context.
