Agentic AI is set to transform buy-side trading by moving beyond market analysis to actively support trade execution, says LTX’s Katie Savignano, director of relationship management and Diana Demianczuk, relationship manager.
Speaking to Trader TV, they explain how the technology is evolving into an autonomous trading assistant capable of handling increasingly complex execution workflows, while stressing that human oversight and robust guardrails will remain essential. As AI becomes more widely available, Savignano and Demianczuk also discuss what they believe will differentiate trading firms in an increasingly competitive market and how they see the generation of AI tools helping desks boost long-term productivity and scale operations.
Interview
Josephine Gallagher – Welcome to Trader TV, where we look at the key issues shaping investment and trading decisions. Today, I’m joined by Katie Savignano and Diana Demianczuk at LTX to discuss how agentic AI is moving from being largely used in market analysis to now being used in real-time trade execution, and what that means for the future of buy-side desks. Katie, Diana, thank you for joining us.
Katie Savignano – Thanks so much for having us, Jo.
Diana Demianczuk – Nice to meet you, Jo.
Josephine Gallagher – So, AI and agentic AI are already being used for data crunching, market analysis, and pattern recognition. What stage is it now at in being able to be used for trade execution, and how do you see it changing the way buy sides operate?
Katie Savignano – Absolutely, as client demand grows, we’re transitioning AI from generating insights, generative AI, to enabling action, agentic AI. For those who aren’t aware, agentic AI involves autonomous software systems that act independently to achieve high-level goals. At this stage, we’re transitioning AI from generative to agentic AI to support trading workflows by monitoring markets, portfolios, and trade blotters in real time, surfacing opportunities, and then acting on those opportunities if the conditions are right, while keeping the trader in control of key risk decisions and final trade approval. This evolution is bringing agility to trading desks, pairing AI’s speed and scalability with human judgment and oversight.
Josephine Gallagher – Excellent. So, can you share some like concrete examples of where agentic AI is actually being applied to trade execution today?
Diana Demianczuk – Sure, agentic AI can detect pricing and liquidity signals, construct trades by filling in trading tickets, select appropriate counterparties and other criteria, stage trade tickets, and even auto-execute based upon trader-defined parameters. The trader can instruct the agent of its task, and that agent will evaluate the market and identify and rank opportunities. The agent then provides a rationale for the opportunities it’s found and auto-populates or auto-executes the trade ticket. So, for example, a trader is looking to buy 10 year triple B plus telco bonds that are trading wider than 100 where their dealers are selling 10 wider versus yesterday’s close. A trader can create an agent that will monitor axe data and actual trading activity to surface relevant opportunities as they emerge, and when specific criteria are met, it can also auto-populate the OIC ticket and alert the trader.
Josephine Gallagher – So, confidence in the output of these agentic AI tools is a big concern that exists today. So, how should guardrails or security measures or validation methods be built into the design of these types of agentic AI tools?
Katie Savignano – Oh, yes, earning confidence in agentic AI, especially in high-stakes environments like trading, depends on designing for controlled autonomy, not full independence. The goal isn’t to remove humans from the loop, but to ensure that AI operates within clearly defined guardrails. Human-in-the-loop approvals for key decisions, policy-based limits on trade size and scope, built-in explainability before any action is taken, and full auditability, so every output and action is traceable and reviewable. At LT, that trust is reinforced by our patented show-your-work approach, which is designed to make AI-driven outcomes more transparent and explainable. It’s also grounded in the fact that over the last three years we’ve built a foundation of established AI tools that have already been embedded in the trading workflow, proving value through compliant productivity-enhancing outcomes over time.
Josephine Gallagher – Excellent. So, Diana, another concern is the proliferation of these agentic AI tools, especially as they become cheaper to build and buy. So, if everybody has an agentic AI tool, how can buy size differentiate? How can they remain competitive?
Diana Demianczuk – As AI becomes accessible and cost-effective, many firms will have access to similar underlying capabilities. As a result, competitive advantage shifts away from the technology itself and toward how it is applied. The real edge comes from proprietary workflows, data, and institutional knowledge. Buy-side firms have deeply valuable differentiated inputs such as execution history, portfolio context, counterparty relationships and internal research that just cannot be easily replicated. A firm specific approach when supported by a trusted AI system allows a trader to produce insights and actions that are uniquely aligned with how that desk actually operates. Then it becomes far more than a generic tool. At LTX, we have years of experience building and deploying AI and corporate bond trading, which gives us a level of expertise and product maturity. In a market where firms claim many similar AI capabilities, maturity and usability become the real differentiators. Buy side-firms will get the most value from solutions that are already proven, embedded in the workflow, and capable of adapting to the way their desks trade. The edge comes not from having AI for its own sake, but from putting it to work in a way that is both proven and intuitive.
Josephine Gallagher – Understood. And now, looking ahead, what do you see being the next phase of development for agentic AI tools in trading?
Katie Savignano – Yes, the next phase of agentic AI in trading is turning tools like Bond GPT into true trading assistance, something that can take on delegated work, monitor continuously, and help move tasks forward without constant manual input. Rather than just responding to prompts, these systems will increasingly function as dependable helpers alongside traders throughout the day. That matters because most firms are under pressure to do more without adding head count. In that environment, AI is not just a productivity tool, it’s a force multiplier for the desk. By taking on repeatable lower touch work, BondGPT can help traders cover more ground, respond faster, and spend more time on the highest value decisions where human judgment matters most.
Josephine Gallagher – Excellent. Thank you very much, Katie and Diana, for joining us.
