Faster-moving liquidity shocks and increasingly interconnected risks are putting pressure on banks to act sooner. For treasury and risk leaders, the speed at which they can identify, assess and respond to emerging threats is becoming almost as important as the risk measures themselves.The shift from reactive, retrospective monitoring to proactive, continuous decision-making was a central theme at Risk Live Europe 2026, where risk leaders explored two sides of the same problem: how banks can respond faster to rapidly developing risks, and where governed artificial intelligence can help them do so without sacrificing control and explainability. Rob Mackay, Regnology In an asset-liability management (ALM) Leaders’ Network session and a separate AI roundtable hosted by Regnology, practitioners discussed how risk frameworks need to evolve. Shortening timeframes were a particular concern in the discussion around liquidity. Real-time payments, 24-hour access to deposits and the rapid spread of information through social and financial media mean liquidity pressures can gather momentum within hours.Traditional measures such as the liquidity coverage ratio remain fundamental, but participants argued that they need to be complemented by intraday monitoring and much shorter stress horizons. Repeated 24- or 72-hour exercises can help banks understand what might happen during the crucial early stages of a crisis and expose weaknesses in their ability to respond.As one participant asked: “What are you going to do in the first 24 hours?” Putting it into actionAnswering that question requires a granular understanding of the balance sheet. Deposit insurance does not necessarily prevent customers from moving money when confidence deteriorates. Banks need to understand concentrations and behavioural characteristics within their deposit base, where liquidity could become trapped and how readily funds can be moved between entities and jurisdictions.You can have a beautiful contingency funding plan, but can you actually execute it?Operational preparedness is key. A written contingency funding or recovery plan provides a foundation, but participants emphasised the practicalities of executing it under pressure. Decision-making authorities, legal attestations, collateral mobilisation, secured funding arrangements, escalation routes and communications may all need to be prepared in advance.“You can have a beautiful contingency funding plan, but can you actually execute it?” said one practitioner.The same thinking applies to stress-testing. Liquidity rarely deteriorates independently of other risks. Capital problems can undermine market confidence, and falling asset values can affect collateral and funding capacity, while reputational concerns can accelerate deposit withdrawals. Participants discussed bringing credit, market, interest rate and liquidity effects into more integrated exercises, including the use of stochastic modelling alongside deterministic scenarios.These pressures strengthen the case for a more continuous approach to risk management. Banks need information sooner, they need to be able to explore scenarios quickly and they need a clearer view of the operational consequences before decisions are made.A practical role for AIThe Regnology-hosted roundtable considered how AI could support that shift. Participants focused particularly on the practical uses of large language models (LLMs) and the controls required before banks become operationally dependent on them.Some of the clearest opportunities sit around existing risk calculations. AI can help ingest information from disparate systems, map data models, clean and validate inputs and reconcile inconsistencies. It can also automate repetitive processes, help users interrogate results and assist with reporting and commentary.Participants were more cautious about allowing probabilistic models into core calculations. ALM valuation, discounting and cashflow engines depend on reproducible mathematical rules. LLMs can produce different answers according to prompts and context, hallucinate information or provide apparently convincing reasoning that does not follow the required calculation methodology.Scenario generation offers a useful example. An AI-generated scenario can appear credible while overlooking an important transmission channel, such as the effect of a yield-curve shock.The discussion therefore favoured purpose-built agents operating within a controlled environment. Deterministic engines would continue to perform the calculations, while AI could organise and interrogate the underlying information, present outputs and support analysis.Suggested safeguards included model validation and backtesting, version control, governed prompt or skills libraries, and traceability of the information used by the model. Human review and approval would remain part of workflows affecting material risk decisions.The agent does the work. You keep the decisionRob Mackay, RegnologyConnecting data, decisions and reportingRegnology Risk Hub shows how this model can work in practice. It is built on Ascend, the firm’s software-as-a-service-first, agentic-first platform – powered by Regnology Intelligence (RGI), a governed intelligence layer that combines intelligent assistance, contextual explanation and conversational scenario modelling with established risk engines and human-in-the-loop oversight.The aim is a unified view of the balance sheet across ALM and interest rate risk in the banking book, liquidity and funds transfer pricing.The approach starts with governed, lineage-tracked data. RGI Explain helps users investigate why a number has changed and trace the movement back towards the underlying positions. RGI Assist supports scenario analysis, enabling users to describe an event in plain language and translate it into a runnable stress scenario, with human approval required where assumptions are material, confidence is low or a final decision needs to be made.For Rob Mackay, chief executive officer of Regnology, retaining that control is central to practical deployment. “The agent does the work. You keep the decision,” he said.Regnology also sees an opportunity to connect risk analysis more closely with finance and regulatory reporting. If decisions and reports draw on the same governed data, banks can improve consistency, reduce reconciliation and strengthen traceability from source data and business decisions through to regulatory reporting outcomes.RGI can assist in drafting explanations or responses to regulatory queries, while approval remains with the user. A shared data foundation addresses a long-standing difficulty for banks whose ALM, liquidity, finance and regulatory reporting processes have developed across different systems. AI layered over fragmented or poorly governed information will do little to resolve the underlying problem.Mackay described ungoverned AI in ALM as “a new model risk”, reflecting the importance the firm places on lineage and explainability. His suggested test for banks assessing AI propositions is similarly straightforward: “Can you explain it to my regulator?”Towards continuous risk managementA fully converged view across risk, finance and regulatory reporting remains a direction of travel for much of the industry. Legacy infrastructure, fragmented data and organisational silos mean many institutions are some distance from it.The discussions at Risk Live Europe nevertheless suggest that the pressure to close those gaps is increasing. Liquidity shocks can unfold faster, risks interact across traditional boundaries and management teams need to understand potential consequences sooner.AI offers tools to reduce manual data work, accelerate scenario analysis and make complex risk information easier to interrogate and act on. Its usefulness will depend heavily on the infrastructure around it: governed data, reproducible calculations, clear decision rights, human-in-the-loop oversight and an audit trail capable of explaining how an answer was reached.For risk and treasury teams facing an increasingly compressed decision-making window, those foundations could determine how far AI moves beyond experimentation and whether it becomes part of day-to-day risk management.
Risk management at the speed of the market
Full Article
Original Source
Read the full article at Risk →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.