For much of the past decade, the conversation about technology in fixed income trading centred on which vendor to select and how long integration would take. That conversation is changing. The cost of building is falling, the speed of delivery is accelerating, and firms are beginning to ask a genuinely different question: why buy what we can now build ourselves?
The shift is being driven by artificial intelligence-assisted development tools. Coding assistants or agent – which can generate, explain and debug code at speed – are compressing development timelines and lowering the technical barriers that previously made in-house software development uneconomic for all but the largest institutions. Several panellists at the Fixed Income Leaders Summit described their technology teams as being measurably more productive, with one noting that the cost of customisation had dropped to a point where the traditional 70/30 build-versus-buy calculus was being fundamentally rethought to 90/10.
The implications for the vendor landscape are significant. If firms can build bespoke solutions faster and more cheaply than before, the case for long-term vendor contracts weakens. One speaker from a major financial institution made the point directly: before committing to a vendor, a firm should now ask whether the capability is something it could develop internally within a reasonable timeframe, and price that option into the decision.
Vendor relationships are not disappearing, but they are evolving from multi-year commitments toward more modular, shorter-duration engagements. As one sell-side speaker noted, “The effect of AI on software-as-a-service vendors is easily observable in their share price.”
That said, a note of caution ran through the discussions. The proliferation of internal tools creates its own risks — fragmentation, governance gaps, and the risk of building something only one person understands, an outcome several speakers compared to undocumented Excel spreadsheets which proliferated in the late 1990s onwards, populated with macros that were fundamental to a trading desk’s management. The answer to tackling this fragmentation most agreed, is discipline around what gets built in-house and what gets bought, with commoditised functions outsourced, and differentiated capabilities retained.
The role of the data scientist within financial institutions is itself in transition. The arrival of AI tools had shifted data scientists away from writing SQL queries toward supervising model outputs, curating training data and acting as AI product managers.
Across the industry more broadly, the boundaries between trading, technology and product functions are blurring. The most effective implementations described were those in which traders, portfolio managers and developers collaborated from the outset – sharing a common platform, a common problem-statement and, increasingly, a common language for discussing what the technology needs to do.
©Markets Media Europe 2025
