UBS wins the global award for best use of AI/ML in FX for embedding the technology into the everyday work of its trading, sales and risk teams. Its programme stands out for the range of production-focused applications, including conversational options analytics, adaptive electronic pricing, complex derivatives valuation, client intelligence and automated sales materials. These capabilities are supported by common infrastructure and a governance framework designed for the demands of institutional FX.
The bank’s approach begins with a focus on practical problems. Its road map during the review period concentrated on execution quality, trader productivity, client responsiveness and scalable risk management. Instead of building a separate innovation layer, UBS placed AI and machine-learning tools inside established platforms and workflows. That reduced the need for users to move between systems and made the technology part of everyday decision-making rather than an occasional specialist resource.
“We’ve always viewed AI as a way to augment human expertise and help us deliver greater value to clients,” says Adrian Boehler, global head of macro distribution. “While it can improve speed and efficiency, its real value lies in helping our teams make better decisions. By turning vast amounts of information into actionable insights, AI enables us to deliver more relevant solutions, better outcomes and a stronger client experience.”
From options analytics to electronic pricing
Ask Neo FX Options illustrates this emphasis on usability. The live service gives traders an always-available assistant within the Neo platform, through which they can ask options questions, retrieve analytics and move from an idea towards pricing. Its significance lies in its integration with the environment traders already use, making it easier to reach relevant information during the pricing process.
UBS is applying a similar principle to electronic execution through Next-Gen eFX Client Pricing. This machine-learning initiative is designed to make pricing more responsive to liquidity, market conditions and the characteristics of client flow. It seeks to balance several commercial considerations, including spread quality and fill rates, and risk.
The bank is also using deep learning to address computational demands in complex products. Its X-Pricer is intended to increase the speed and scalability of valuation and risk calculations for FX exotics. The underlying pricing environment includes EQSVal and UBS pricing libraries, while Neo provides the interface for pricing, analytics and conversational AI.
Building for adoption, control and client relevance
A shared technology foundation gives the programme greater coherence than a collection of desk-level tools. UBS uses Nucleus and AI Mart for standardised machine-learning operations, alongside RiskLab capabilities covering model lifecycle management, testing, deployment and monitoring. Large language models used by trader-facing applications are deployed behind the bank’s firewalls, with the bank emphasising confidentiality, resilience and low latency.
This architecture supports repeatable deployment. Central platforms such as AI Mart and AI Model Ops provide shared infrastructure, governance and deployment standards, while embedded specialists connect those foundations to individual use cases. Operational platforms such as OnePlace extend the same approach into know-your-customer and workflow processes.
While [AI] can improve speed and efficiency, its real value lies in helping our teams make better decisions
Adrian Boehler
The benefits extend beyond the trading desk. Client 360 AI is designed to provide sales teams with stronger client context. Automated FX hedging presentations can turn live pricing into client-ready material in seconds, replacing a more manual-based production cycle. In combination with Ask Neo FX Options and the electronic-pricing work, these tools aim to shorten the path from analysis to a client conversation and from a hedging idea to a usable proposal.
UBS’s strength is shown when these applications are considered as connected parts of the client service model. A trader can retrieve relevant analytics sooner; a salesperson can enter a discussion with more context; and a hedging conversation can progress more quickly from live pricing to a presentation. In electronic channels, machine learning is being used to refine the balance between execution certainty, spread quality and risk. AI therefore supports several points in the client’s experience with the bank: pricing, analysis, communication and execution.
The programme’s governance is equally important. UBS applies a risk-tiered framework that distinguishes between narrow AI and agentic and generative AI systems. Requirements are calibrated according to factors including model risk, explainability and runtime behaviour. For higher-risk applications, the framework places greater emphasis on guardrails, monitoring and identified sponsor and owner accountability.
Controls include model validation and testing, explainability techniques, performance monitoring and drift detection. Trading algorithms, including the Next-Gen eFX pricing engine, have fail-safe mechanisms such as kill-switch triggers. Human-in-the-loop oversight retains accountability for critical trading and risk decisions with traders and risk managers. The framework is aligned with best execution obligations and the FX Global Code, connecting technical governance with the conduct requirements of the market.
UBS’s people model reinforces this structure. Traders, eFX quants, sales and operations specialists, risk managers, data scientists and engineers contribute to development. Embedded AI specialists work directly with front-office and risk teams, while a network of AI champions across FX supports adoption. Training covers group AI policy, model governance, monitoring and broader AI learning journeys.
UBS wins Euromoney’s global award for best use of AI/ML in FX because it has joined these elements together. Its AI capabilities address identifiable needs, operate within core workflows and are supported by common infrastructure, specialist teams and explicit safeguards. The bank is not presenting AI as a substitute for market expertise. It is using it to make that expertise easier to access, faster to apply and more consistent at scale.
