By Carl Thornberg, head of optimisation and analytics technology, OSTTRA.
In post-trade operations, most problems are not caused by outright errors, they are caused by ambiguity. Trades that look different but are not wrong. Valuations that diverge for valid reasons. Numbers that do not line up, even though nothing has actually gone awry. At scale, that ambiguity is more than a nuisance – it becomes very expensive.
Across global markets, banks reconcile vast portfolios of trades every day. In most cases, counterparties broadly agree. But a small proportion of differences persist, feeding into portfolio reconciliation breaks, valuation discrepancy and, ultimately, collateral disputes. Each instance requires manual intervention, devouring time, human attention and in many cases, regulatory capital.
This is where artificial intelligence (AI) is beginning to matter in a tangible, impactful way.
A dispute problem, not a broken system
It is important to be clear about what this problem is and what it is not. Post-trade infrastructure is not failing. On the contrary, trades are confirmed, processed, and settled at extraordinary scale with remarkable reliability.
The challenge emerges much later. Over time, trades that once matched perfectly can appear differently in each counterparties’ internal systems. Present values move as markets move. Models diverge, volatility assumptions vary and FX rates are captured at different times of day. Time zones, calendars, and internal conventions all play a role. While most of these differences are often valid, proving that is difficult.
As a result, banks often devote large teams to dispute management. Dozens of people may spend their days drilling down from portfolio-level differences to individual trades, trying to answer one deceptively simple question: ‘why do we disagree?’
The real risk is not that differences exist. It is that genuinely dangerous booking errors can be hidden among a much larger volume of explainable noise.
Signal versus noise
This is the distinction that really matters. In dispute management, the signal represents the handful of true errors that can pose genuine financial risk. The noise is everything else: timing effects, model differences, data conventions and benign inconsistencies that look alarming until properly explained.
Historically, separating the two has been slow and manual. Teams work through disputes one by one, often without the full context needed to resolve them quickly. The result is operational drag, capital buffers held “just in case”, and less time spent on the issues that genuinely deserve attention.
AI changes this dynamic, not by replacing expertise, but by accelerating understanding.
What AI actually does in this context
The value of AI in post-trade is not abstract. It lies in pattern recognition across scale. Post-trade platforms occupy a unique vantage point maintaining a view of both sides of a trade. This means that not only do they witness how valuations evolve over time, but crucially – how similar disputes have been resolved in the past. Individual institutions simply cannot replicate this view on their own.
By applying advanced analytics and AI to this dataset, it becomes possible to explain a far greater proportion of differences automatically. Not by guessing, but by learning from history.
For instance, valuation differences driven by FX timing can look like serious breaks when viewed in isolation. But when analysed across time series data, exchange rate movements and historical behaviour, they can often be identified and explained with high confidence. What once required hours of manual investigation can be resolved far more quickly, and with clear supporting evidence.
From investigation to prioritisation
When explainable differences are resolved faster, two things happen. Firstly, operational teams spend less time proving that nothing is wrong. That reduces cost and friction across reconciliation and collateral processes.
Secondly, and more importantly, with the noise filtered out, the remaining pool of unexplained differences stands out more clearly. This is where genuine booking errors, model failures, or contractual misunderstandings hide.
In other words, AI helps teams prioritise risk, not obscure it. Adding to the toolbox, not replacing it. None of this suggests a radical break from existing post-trade practices. Human judgement remains essential – but what AI adds is leverage. It enhances the existing toolkit by removing friction and ambiguity at scale. It also allows experienced professionals to spend more time on high-value work and less time navigating false positives. This is particularly important as volumes continue to grow and markets become more interconnected. Complexity is not going away. The only sustainable response is better insight.
A pragmatic path forward
AI in post-trade does not need to be futuristic to be transformative. Its impact is already visible in dispute explanation, reconciliation efficiency and collateral workflows. The next phase is about extending that capability responsibly – applying intelligence where data is rich, outcomes are measurable, and human decision-making is enhanced, rather than displaced.
The goal is clarity, less noise, and sharper signals. By ensuring risk is no longer drowned out by ambient friction, AI facilitates a post-trade environment where material exposure is easier to identify and manage. While AI adoption is in its infancy, momentum will build as firms realise tangible, measurable gains in operational efficiency.
Subscribe to our newsletter
