Fixed income leaders are struggling to execute their AI ambitions, not for lack of enthusiasm, but because the data foundation underneath those ambitions isn’t ready.
This was the main topic that emerged at Clearwater’s roundtable discussions on workflow and data, at the Fixed Income Leaders’ Summit 2026 in Boston with buy-side fixed income traders, portfolio managers, and trading technologists from large asset managers, pensions, and insurers.
What emerged from the conversations wasn’t a shortage of enthusiasm for AI, but clarity about the data gap. Firm after firm articulated, in its own way, that the data foundation underpinning AI strategy is not there yet. For companies, the cost of not solving these problems is rising.
This was the loudest theme across every table. Investment data lives in silos across legacy platforms, vendor systems, proprietary tools, and spreadsheets, and it doesn’t move cleanly between them. Downstream systems were configured years ago and don’t carry the fields that newer asset classes require, so teams patch the gaps manually.
Nobody owns the data definitions or the lineage, so when a number looks wrong there’s no clear place to go. A recurring example is trading desks need one trusted real-time exposure figure to size orders against. This is built on ratios, hedges, and derivative-adjusted positions. Instead of one clear place to go, there are several versions of that number living across the OMS, the risk system, and accounting, with no reliable way to determine which is the truth.
The attendees found the problem isn’t a shortage of data, rather a shortage of connected, governed data.
Traders, PMs, and analysts across these firms are building their own tools, dashboards, and agents using ChatGPT, Claude, and Copilot. The capability is real. The cost is proliferation.
People who have never written code are now writing Python. The result is a growing pile of ungoverned, unshared, and unvetted applications. Someone builds something genuinely useful, and there’s no infrastructure to hand it to a colleague or scale it across the firm. Tools live everywhere. Nothing is governed, and there’s no golden source of truth underneath any of it.
A few shops are putting structure around this, an AI development lifecycle, a registry of agents, but every firm said they’re in the early stages, and nobody has claimed to have solved it.
More data has not produced better decisions, and the room was clear-eyed about why that’s dangerous when AI is doing the reasoning. The risk isn’t a system that says it doesn’t know. It’s a clean, confident answer pulled from data that isn’t reconciled or fully understood.
One example made it concrete. Trading volume returned a figure five times too high because the AI summed cancels and corrects without understanding what they were. The point that resonated across the room is that reconciled, well-understood data matters more in the AI era, not less. You can’t point a model at unreconciled, replicated data and trust what comes back.
An underlying thread ran alongside this: where AI has earned the right to act versus only to suggest, and what conditions a desk would need confidence before trusting a system to execute rather than recommend.
The most concrete and demonstrable pain in the room was that vendor pricing and model analytics fall down on short-duration and near-maturity securities, less-liquid names, and securitized products. Prices go stale and don’t refresh, distorting yields. Durations come wrong on structured products.
The workaround everywhere is manual processes. A trader overrides by judgment, and that lives in a spreadsheet or an aging database. It doesn’t scale, and it doesn’t get better on its own.
The throughline across every conversation at FILS was the same. The AI ambition is real, but the data foundation to support it isn’t there yet. Fragmented systems, ungoverned tools, unreconciled data, and manual workarounds on the hard-to-value book are not peripheral complaints. They’re the thing standing between where these firms are and where they want to go.
Clearwater’s front-to-back, reconciled data foundation is built for exactly this moment. A single source of truth that front office, risk, and accounting all work from. A governed environment for the analytics and workflows currently scattered across spreadsheets and one-off tools. Data validation and override capabilities for the pricing gaps that every fixed income desk is papering over manually today.
The firms that will get the most from AI aren’t the ones with the most ambition. They’re the ones with the cleanest foundation underneath it. That’s what Clearwater provides.