Humans still in charge, for now
Within a couple of years, Wall Street banks may have AI agents executing bond trades, but seasoned traders will still need to supervise and guide the process to avoid drift. That view is shared by Michael Roberts, the executive running HSBC Holdings Plc's corporate and institutional banking division, who recalls recent instances when bond-trading algorithms stumbled during bouts of market stress. Hiring is already shifting in response, with more AI fluent juniors and additional quants coming in to build and manage these tools.
Investors are bracing for the change. "Based on all the conversations that we've had with clients, everybody's preparing for this," said Zornitsa Todorova, who leads thematic fixed income currency and commodities research at Barclays, about bots taking on more execution. "It's a slow burner, but it's definitely going to happen."
What AI is doing on desks today
Right now, most agents act like research assistants. Market participants say they are pulling data, parsing complex bond covenants and condensing headlines to support pre trade decisions. They are also building screeners and visualizations and automating internal chores.
Izzy Conlin, who leads strategy and solutions across Tradeweb's global markets, said the firm is seeing "a ton of interest" in applying AI to pre- and post-trade analytics, cost comparisons across protocols, dealer selection for specific orders, and assembling portfolio trades with appropriate bonds. Even with that momentum, she added, "I still don't see a world where you don't have human traders," noting that trying to quickly change a decade of dealer and investor habits "doesn't work."
Platforms push adoption, with limits
Venues are embedding AI into trading workflows. In June, LTX added agentic features to its BondGPT application. The upgrade, called Agentic BondGPT, can take predefined steps such as drafting trade tickets or initiating trades for a trader. "AI agents are no different to hiring a human trading assistant on the desk," said LTX Chief Executive Officer Jim Kwiatkowski.
Agents can also surface ideas in the harder to reach corners of credit by flagging price quirks or finding value in the intricate paperwork of distressed companies that have been restructured. A typical rule is to notify the desk if a CCC-rated bond widens by over 50 basis points relative to its previous day's closing level. Additional triggers include highlighting high-yield bonds that shift by 1 point up or down compared with the prior day, and auto-populating a ticket whenever there's an offer of at least $5 million on a 10-year industrial bond with a BBB rating or higher and minimum spread of 85 basis points.
When market technology keeps advancing, staying clear on what matters becomes more important, so sign up for the free Market Briefs
Even with those assists, traders say outputs need human checks, and most are not allowing agents to execute without oversight. In some cases, compliance rules prohibit using AI agents at all. "We view it as a tool that augments human decision-making rather than replaces it," said John Lloyd, Janus Henderson's global chief for multisector and corporate credit.
Why it matters for markets and your portfolio
Liquidity matters. On a typical day, about $62 billion of US corporate bonds change hands, according to Crisil Coalition Greenwich. As trading digitizes, Barclays sees investment grade electronic activity rising toward 65% of volume from roughly 45% today, while high yield could reach about half of trading. Tradeweb, already the No. 1 US corporate bond e trading platform, is purchasing Trumid, ranked third, for about $1.97 billion to capture more of that flow.
There are risks to watch. Barclays warns that if many models respond to the same signals at once, they could produce "episodic volatility" across credit. Consider the backdrop: since late September, the average spread on CCC rated US corporate bonds has been above 1,000 basis points, a level that usually points to high odds of default, restructuring or loss. In plain English: more automation can speed discovery and execution, but humans are still set to override the bots when the tape gets jumpy.
As trading tools evolve, a simple daily check can save time and confusion, so try the free Market Briefs newsletter
