(Image source: IIR/Gemini.)
Financial institutions have long used internal data to monitor risk, but volatile markets, digital channels and emerging exposures are creating demand for broader early warning capabilities. For banks, that can include balance sheet modeling, liquidity-risk monitoring and external signals that may correlate with commercial and retail deposit behavior. For insurers, the relevance lies in combining core operational data, advanced analytics and external signals to improve risk sensing.
Banorte’s (Mexico City) work with SAS (Cary, N.C.) technology illustrates that pattern. The Mexican financial group is using SAS ALM and SAS Viya as part of a broader risk analytics program that includes balance sheet modeling, interest-rate risk management, climate-risk integration and, more recently, a liquidity-risk initiative connecting core banking data with social-media sentiment analysis. Using natural language models and in-house machine learning, Banorte is developing early warning indicators intended to help risk managers correlate public sentiment with deposit behavior.
IIR met Abraham Izquierdo, Managing Director of Treasury Risk, Banorte, at SAS Innovate 2026 in Grapevine, Texas, where SAS introduced him as a customer whose work in banking risk management had relevance beyond banking. Izquierdo described how Banorte is using SAS technology, core banking data and in-house machine learning to connect liquidity-risk monitoring with external sentiment signals.
“SAS has been a strategic partner for us in many fronts,” Izquierdo says. “One of those is risk management.”
From Balance Sheet Modeling to Broader Risk Analytics
Banorte’s AI-enabled liquidity-risk initiative builds on a broader risk analytics program that has included balance sheet modeling and interest-rate risk management, areas where the institution has used SAS technology to model banking-book positions and analyze the behavior of assets, liabilities and hedges.
Banks must understand how interest-rate changes affect the banking book, which requires modeling a wide variety of balance sheet positions, including products with optionality, early redemption behavior and non-maturing deposits. Banorte has used SAS ALM for modeling banking-book positions, characterizing products with optionality or early redemption, estimating the expected duration of non-maturing deposits and drawing conclusions about natural hedges or duration hedges.
The earlier balance sheet modeling work had a practical business purpose: proper measurement and management of interest-rate risk, with tangible effects on financial margin. Through the modeling, Banorte could better understand the behavior and rate sensitivity of balance sheet components over a proper time frame. The firm also found that natural hedges can, in some circumstances, substitute for financial derivatives with explicit costs.
The more recent liquidity/sentiment project reflects a continuation of that discipline: bringing more data, more modeling and more analytical sophistication into established risk-management functions. For insurers, the particulars differ, but the familiar challenge is translating risk drivers—whether interest rates, climate, claims volatility, social inflation or cyber accumulation—into models, indicators and management actions.
Connecting Liquidity Risk and Sentiment
Within the newer initiative focused on liquidity risk and deposit behavior, Banorte has connected its SAS liquidity-risk capability with the bank’s core banking platform and social-media data, using public comments from X, formerly Twitter, as an external sentiment source.
“This initiative is connecting our SAS tool for liquidity risk with our core banking and with social networks,” Izquierdo says. “This has strengthened our relationship with our strategic partner, SAS.”
The business context is commercial and retail deposits. The purpose is to establish early warning indicators by correlating deposit behavior with public sentiment. Positive sentiment may indicate no immediate concern, while negative sentiment may warrant further analysis to determine whether it corresponds with changes in deposit patterns.
“The linkages try to establish early warning indicators by connecting this measurement of liquidity risk based on opinions on social networks,” Izquierdo says.
The AI component has two related roles. Natural language models translate public social-media comments into sentiment indicators, while machine learning is used to determine which correlations best fit the relationship between social-media reactions and deposit behavior.
“You have to use natural language models in order to translate whatever you say in the social network, in social media and then translate into an indicator which we might correlate with our deposit base, our deposit behavior,” Izquierdo says.
The machine learning model learns from historical numbers and continues learning as more data becomes available, Izquierdo adds.
“The other important role is that you use a machine learning model in order to segregate and to order which correlation adapts better between these social media reactions and our deposits behavior,” he says.
Banorte began the initiative roughly a year ago and had been in production for about two months at the time of SAS Innovate 2026. Izquierdo characterizes the model’s early performance as “behaving correctly.”
Extending the Field of Vision
The liquidity/sentiment initiative extends, rather than replaces, established risk-management disciplines. Social-media sentiment does not make risk decisions; it becomes an additional signal that can prompt further investigation when it appears to correlate with deposit behavior.
Insurers face comparable analytical challenges in areas such as claims, underwriting, fraud, litigation, weather exposure, customer behavior and reputational risk. The sources differ, but the discipline is similar: connect internal operational data with external signals, identify meaningful relationships and keep risk professionals in the interpretive loop.
Built In-House on SAS Technology
Izquierdo distinguishes between the SAS technology foundation and the model Banorte has built around it. The liquidity-risk piece runs on SAS Viya, while the broader risk-management relationship includes SAS ALM. The sentiment/liquidity model itself was developed internally.
“We used the SAS tool and our core banking tool, but the model itself was made in-house,” Izquierdo clarifies.
Large financial institutions may rely on major technology platforms for infrastructure, analytical tools and model-management capabilities, but the most valuable use cases often depend on proprietary data, internal expertise and domain-specific judgment.
For Banorte, that means the risk organization needs specialized talent. Izquierdo says AI is evolving rapidly, and institutions need people with the skill set to implement and train models within specific business functions such as risk management.
The project’s value, in his view, is strategic. Izquierdo says the liquidity/sentiment initiative can become an important tool for the group and a starting point for other AI initiatives within his team.
“I hope one important expected result is that this becomes a strategic tool for the group and a pillar to expand to more applications related to AI,” Izquierdo says.
Centralizing AI Around Business Strategy
Izquierdo also points to a broader governance issue: how large financial institutions prevent AI activity from becoming fragmented. One of the major challenges, he says, is that AI initiatives can develop in silos, with different teams working with different consultants or AI companies without sufficient coordination.
“One important challenge for many institutions is that AI has to be centralized,” Izquierdo says. “Because if not, that silo over there talks to one consultant, to one AI company and the other also. And in silos, they work with different initiatives.”
Banorte’s approach includes a centralized AI team that aligns initiatives with group strategy, according to Izquierdo. That centralization is not presented as a barrier to innovation, but as a way to ensure that innovation contributes to enterprise goals.
That view is consistent with comments Izquierdo contributed to a SAS banking AI thought-leadership e-book, where he emphasized leadership, culture and pacing. “Innovation and AI must be recognized as a pillar of the strategy in the institution,” he comments. “It’s important to ensure leadership understands the opportunity and sees the results.” In the e-book he also stresses that AI and cloud deployments should be “multi-year and step by step,” with short-term goals, compliance and trust among senior stakeholders.
For insurers, the governance question is increasingly familiar as AI adoption moves from individual pilots toward enterprise implementation. The practical issue is whether organizations can govern, integrate and scale AI in ways that support business strategy, regulatory expectations and operational trust.
Climate Risk and Business-as-Usual Integration
Banorte’s climate-risk work applies the same risk-management discipline to a different category of emerging exposure. Through its Sustainable Strategy Plan, the company has developed a climate transition strategy that includes modeling expected and stressed physical and transition-risk impacts under multiple scenarios and incorporating climate considerations into systems, processes, pricing, governance and stakeholder reporting.
The work brings climate stress testing into ordinary risk management. Banorte evaluates physical risks such as cyclones, flooding and storm surge, as well as transition risks tied to carbon-price pathways and greenhouse gas emissions. The objective is to translate climate scenarios into measures of potential credit and financial impact, rather than treating climate risk as a separate sustainability exercise.
The connection to insurance is direct. Physical and transition risk already affect underwriting, pricing, reserving, capital management, reinsurance and enterprise risk management. Banorte is approaching the problem from the standpoint of credit and balance sheet risk, but the analytical discipline is similar.
Extending the Risk Discipline
Izquierdo does not quantify the cost or financial return of the liquidity/sentiment initiative, but he characterizes the expected value as substantial relative to the investment.
“The potential benefit is great and very relevant for the group,” he says.
He also sees the initiative as a starting point for further AI work within his team. Its value may lie not only in liquidity-risk monitoring, but in the way it extends an established risk discipline with new signals, internal modeling and centralized governance.
For insurers, the particulars will differ, but the pattern is familiar: emerging risk often appears first at the edges, in signals that have to be connected to core business data before they can inform decisions.

