Application of Gen AI to mitigate Treasury Risk
As the business adage goes, we need to take proper care of our credit to ensure that our credit takes good care of us! With this in mind, credit risk, also known as default risk, is the risk inherent within financial transactions that occurs due to the probability of counterparty default on the financial exposure. The potential for financial loss arises when a borrower is unable or unwilling to fulfil their contractual payment commitments to a lender or creditor. As well as exploring this theme in the context of the treasury department, this article will highlight the different models used to arrive at a credit score, and consider the nature of credit ratings of counterparties and issuers of financial instruments.
Credit risk can have a significant impact on a treasury department’s operations and overall financial stability. Its function can be impaired by the credit risk of existing and future liabilities, as well as through the management of assets/liabilities, cash and future cashflow, and the handling of financial risk management.
Credit risk destabilise treasury in various ways:
If a company’s creditworthiness deteriorates, it may face difficulties in accessing funds, and in securing short-term financing to manage its liquidity.
Treasuries manage a company’s cash, and invest its surplus funds. Credit risk influences the selection of financial instruments and institutions where cash is held or invested. Lower-rated instruments may offer higher returns but come with greater credit risk.
Treasury departments often engage in financial transactions with other FIs, such as banks, brokers, and investment firms. Credit risk associated with these counterparties can impact the security of investments and the settlement of transactions.
Treasuries make investment decisions regarding the company’s surplus cash. Credit risk affects the selection of investments, particularly when choosing between low-risk assets (e.g. government bonds) and higher-yield but riskier assets (e.g. corporate bonds).
Credit risk can affect the management of accounts receivable and accounts payable. Treasuries must strike a balance between extending credit to customers while managing the risk of delayed or non-payment.
A company’s cost of borrowing is influenced by its credit risk. A higher credit risk may result in higher borrowing costs and reduced access to capital markets.
Treasuries must adhere to regulatory requirements related to credit risk management, such as capital adequacy standards and stress testing.
Monitoring the credit ratings and financial health of issuers, counterparties, and holding companies is crucial for treasury departments. Downgrades or negative credit events can prompt the need for adjustments in investment strategies or risk management.
Treasury departments often provide regular reporting on credit risk exposure to senior management and the board of directors. They assess the potential impact of credit risk on the company’s financial stability and solvency.
Financial rating agencies, such as Standard & Poor’s (S&P), Moody’s, and Fitch, evaluate the credit quality of financial instruments, issuers, and economic ecosystems including country (or sovereign) ratings, in order to place investments on a level playing field.
Specific criteria used to rate financial instruments include:
The treasury department, while investing and integrating the riskiness of the valuation and revenue from the rated instrument, needs to bear in mind the dynamic nature of ratings, which generally deteriorate over time. It should also be mindful of the Five Cs of credit:
A credit score specifically refers to a three-digit expression of an individual or a businesses’ credit background and creditworthiness, including credit history, payment history, credit use, and duration of credit history. This expression is instrumental in decision-making for loan disbursement and states the riskiness of the loan account. Treasury monitors loan riskiness, and hence the expected loss of the loan portfolio.
There is a growing trend across many benchmarks emerging economies where the gap between net financial assets – assets of households versus financial liabilities – is reducing. This is impacting the credit scores of individuals. Treasury should consider this trend when lending out secured and unsecured credits.
Treasurers should also keep an eye on the variables of their risk models, and their correlations, that determine the firm that are on dangerous ground and their distance to default, also commonly refer to Probability of Default (PD). The credit default exposure primarily depends on the probability of an instrument’s default, the amount treasury is exposed to the instrument (Exposure at Default, EAD), and the default severity of the distressed asset that determines the amount that can be regained via auction.
The expected loss is a function of default severity and probability. Moreover, treasurers should make provision to cover for unexpected losses, defined as maximum potential loss or maximum loss at a stated confidence level, such as 95%. Unexpected losses are not incorporated into the default provision, but do require that capital is available to prevent financial failures in the event they do occur.
The following are protective guidelines for treasurers:
By optimising capital allocation, treasury must incorporate capital requirements and capital charges defined by regulators. Economic capital is the amount of capital necessary to act as a buffer against unexpected declines from an asset or business line.It is important to optimise the amount of capital so that the firm can operate efficiently. To optimise, we need to minimising the unexpected losses. The constraint when attempting to find the optimal portfolio is that the total amount of capital should be greater than unexpected losses. This ensures that target Return on Equity (RoE) is met or exceeded during the allocation process. Another widely used yardstick to measure the value-producing capacity of an asset is Risk-Adjusted Return on Capital (RAROC). It’s a function of weekly volatility and the tax rate. The RAROC process uses a worst-case scenario approach to estimate the future value of an asset, then produces the capital cushion necessary to provide for the potential loss.
Some widely used and accepted models pursued by risk team include:
Merton model – This model rides on the Black-Scholes option model used to access the value of a firm’s stocks and bonds and compute the probability of default. It states the value if a firm’s outstanding debt (D) and equity (E) are equal to the value of the firm (V=D+E). In another way, the value of the debt can serve as an indicator of a firm’s default risk.
Moody’s KMV (Kealhofer, McQuown, and Vasicek) model is built on the Merton model and adjusts some of the flaws, most outstandingly that:
More realistic and accurate, but complex to calibrate, models include CreditMetrics and Algorithmics’ Mark to Future (MTF). CreditMetrics uses rating migration probability, which includes upgrade and downgrade, thus making it a marked-to-market model. MTF incorporates elements of credit, market, and liquidity risk and uses numerous risk factors to simulate value.
Basel II, and subsequently Basel III, sets capital adequacy rules for banking institutions. A set of credit risk measurement techniques were proposed under Basel II when it was launched. The Basel Committee on Banking Supervision introduced the Basel II Accord to permit banks a choice between two broad methodologies for calculating their capital requirements for credit risk: Standardised Approach (STA) and Internal Rating Based (IRB) approach. Both methods have a common goal: to calculate the Risk Weighted Assets (RWA) accurately, diligently, and with integrity.
RWA is calculated with a pre-agreed risk weight factor to an asset. It is an inverse function in calculation of capital requirement or Capital Adequate Ratio (CAR) for a financial firm. The objective of the bank is to stack the minimum amount of capital that a bank must possess, depending on the risk charge of the bank’s lending activities, investments, and other assets.
The issuer or source of the asset is most important factor when determining the credit rating. The risk of a sovereign-issued security will always be perceived and allocated the highest rating when compared with a corporate bond or with a riskier corporate commercial lending. However, a Google or Apple share will have a far superior rating than sovereign instruments issued by a financially and politically challenged nation.
Adopting a credit risk measurement approach becomes crucial when determining the CAR and capital cost. The capital cost is in terms of credit loss and opportunity cost to park the capital to cover for the CAR. The motive to select and subsequent supervisory review and market disclosure should be done by intent use regulatory policies to stay afloat. The governance should be comprehensive, transparent, and updated periodically. Although FIs have the flexibility and the option to choose n a method, only those meeting certain minimum conditions, disclosure requirements, and approval from their central bank like Bank of England and ECB are allowed to conduct using the internal ratings-based (IRB) approach in estimating capital for various exposures.
The accurate credit rating of the instrument, therefore, becomes a point of utmost importance. The risk weight to the assets is based on the assigned rating. The migration of the rating due to the impact of derogation of asset should be precisely calculated. The triggers of the 2008 financial crisis included the vulnerability, volatility, unpredictability, and migration of credit ratings of issuer, country, or financial instruments. Miscalculating the instrument rating will always cause a failure to predict catastrophes such as the sub-prime crisis, because the prediction will be based upon on quantities of wrong and misinterpreted information.
A firm’s treasury department should have the governance and capabilities to understand the corporate’s assets and its value along with the associated risk and implications. Treasury should liaise with risk management, middle office, and compliance teams to collaborate on various policies and governance strategies. The liquidity and cashflow management should be sound- and shock-proof with credit risk at the heart of all considerations. Treasurers must have the capability to process complex statistical calculations and permutations in order to use the most accurate and appropriate model, which is being continuously fine-tuned.
The ever-evolving world of finance is propelled forwards by technological innovation. Among the most significant developments of the past few years is the adoption of generative artificial intelligence (GenAI) by the world’s largest FIs.
This game-changing innovation is changing the face of the financial industry and providing unmatched value to businesses in various ways. Some of the world’s largest financial organisations have already begun using GenAI to reimagine traditional risk assessment and credit rating methods. By processing enormous datasets, GenAI dramatically boosts the accuracy of credit decisions. As per the market research, these FIs report up to a 30% improvement in credit risk prediction, reducing default rates and enhancing profitability.
Gen AI-based credit scoring is a relatively new approach for assessing an individual’s creditworthiness that uses large language models (LLMs) in conjunction with AI-driven algorithms to quickly sift through mountains of individual data including digital footprints on online transactions, social media interactions, browsing behaviours. It can also enhance the financial inclusion of new consumers by allowing for the analysis of the creditworthiness of consumers with limited credit history. Thanks to Gen AI-based credit assessment, lenders can make better-informed decisions about the credit risk of borrowers with little or no credit history.
The comprehensive analysis of AI-based credit scoring can be achieved by integrating ML with traditional scoring. These algorithms are trained on large sets of historical data, from which they identify patterns and correlations related to a borrower’s ability or likelihood to repay a loan. The patterns from historical data generate new data to predict future behaviour. This process of learning from past data to make predictions is a primary game-changer of ML, making AI-based credit scoring possible. However, the lenders can gain greater insight and a more complete picture of a borrower’s creditworthiness by utilising advanced ML algorithms and the power of LLMs, which can filter and distil massive volumes of data.
GenAI emerges as a powerful ally in the ongoing fight against financial crime. Financial organisations may save billions of dollars annually using sophisticated algorithms to spot fraudulent activity. A GenAI-powered approach can go beyond static credit scores and includes dynamic insights into how individuals and organisations earn, spend, and save money. FIs can use this type of in-depth evaluation to assess the credit risk of new borrowers such as young people, those who are new to a country, and enterprises without standard credit scores.
The diagram below illustrates the GenAI-driven process flow:
Generative AI will make the work of financial analysts easier by putting all the necessary information at their fingertips. During this procedure, Generative AI will comprehend the query that Financial Analysts have submitted, conduct an intelligent search based on the query, activate the necessary AI/ML component from the AI/ML layer, and then indicate whether the lending request should be approved or rejected.
Credit risk can be ring-fenced with various internal checks and best practices; it’s an inherent and integral part of wherever and whenever a financial transaction occurs.
Regulatory compliance, internal governance, stricter supervision, and leverage of the GenAI-driven compliance checks must collaborate to analyse better, identify, measure, and mitigate credit risk.
Authors:
Bhushan Joshi
Email: [email protected]
LinkedIn: https://www.linkedin.com/in/joshibhushan/
Raja Basu
Email: [email protected]
LinkedIn: https://www.linkedin.com/in/basuraja/
Anjanita Das
Email: [email protected],
LinkedIn: https://www.linkedin.com/in/anjanita-d-6b52368/
Sudip Mukherjee
Email: [email protected];
LinkedIn: https://www.linkedin.com/in/sudeep-mukherjee-9765b014/