Use case #0001

How Stress Testing AI models a 200bps rate shock on your bank portfolio

A 200 basis point rate shock is not a hypothetical — it is recent history. Between May 2022 and February 2023, the Fed raised the federal funds rate by 250 basis points over seven consecutive decisions. For a bank with a fixed-rate lending book funded by floating-rate borrowings, each of those hikes compressed the net interest margin in real time. The Stress Testing Agent AI models what that compression looks like across the specific composition of your portfolio — not the industry average, but your actual book.

Why a 200bps rate shock is the canonical bank stress scenario

The Fed / OCC's supervisory stress testing framework for banks specifies that interest rate risk must be modelled under a minimum shock of 200 basis points — applied instantaneously to the rate curve, not as a gradual shift. This is a severe test by design: an instantaneous 200bps move is larger than any single Fed rate cycle in recent history, but it is not implausible over a 24-month horizon, and the "instantaneous" assumption ensures that the test captures the worst-case impact without the benefit of gradual adjustment.

For a bank, the impact of a 200bps shock transmits through four channels simultaneously: the cost of funds channel (borrowings reprice upward, compressing the spread), the asset quality channel (borrower monthly payment increases or DTI breaches trigger defaults), the liquidity channel (refinancing costs increase and rollover risk rises), and the capital adequacy channel (higher provisions on stressed assets reduce Tier 1 capital). The Stress Testing Agent AI models all four channels from the institution's actual portfolio data — not from a generic template.

"A stress test that models the industry average is a stress test that describes someone else's risk. The Stress Testing AI models your portfolio, your funding mix, and your borrower profile — and produces a result that is actually actionable."

The four transmission channels — and how the AI models each

01
Channel 1 — Cost of funds

Floating-rate borrowings reprice upward — fixed-rate assets do not

The Stress Testing Agent AI reads the institution's current funding book: what proportion of borrowings are floating-rate (NCDs, bank lines, CP), what proportion are fixed-rate (long-term NCDs), and what is the reset schedule for each. In a 200bps shock, all floating-rate borrowings reprice at the new rate. Fixed-rate borrowings are unaffected until their maturity. The AI computes the weighted-average cost of funds before and after the shock, and derives the NIM compression: the change in spread between the average lending rate (adjusted for fixed vs floating asset mix) and the new weighted-average cost of funds.

02
Channel 2 — Asset quality

DTI increases as EMIs rise — borrowers at the margin breach the threshold

For floating-rate loan assets, a 200bps rate increase translates directly into higher EMIs. For a $20 hundred thousand personal loan at 11.50% over 5 years, a 200bps increase to 13.50% raises the monthly payment from approximately $44,000 to $46,200 — an increase of $2,200 per month. For borrowers already at 43% DTI, this increase pushes them to 46% — above the policy ceiling. The Stress Testing AI segments the portfolio by DTI and the additional monthly payment cost at the shocked rate, computing how many borrowers cross the 45% threshold and their probability of default at that new DTI level.

03
Channel 3 — Liquidity and refinancing risk

Refinancing costs increase — rollover risk rises for short-tenor borrowings

banks with significant short-tenor liabilities (commercial paper, short-term NCDs) face refinancing risk when rates spike: they must roll over maturing liabilities at higher rates, which accelerates the cost of funds increase. The Stress Testing AI profiles the institution's liability maturity schedule — what refinances in the next 3 months, 6 months, 12 months — and computes the cumulative cost impact of rolling those liabilities at the shocked rate. Institutions with a significant proportion of short-tenor liabilities (more than 30% maturing within 6 months) face a faster-transmitting shock than those with longer-dated funding.

04
Channel 4 — Capital adequacy

Higher provisions on stressed assets reduce Tier 1 capital — CET1 ratio may fall below regulatory floor

The asset quality deterioration from Channel 2 translates into provisioning requirements. Loans that migrate from Stage 1 to Stage 2 or Stage 3 under the stress scenario require higher provisions under Ind AS 109 expected credit loss methodology. The additional provisions reduce net profit and, through retained earnings, reduce Tier 1 capital. The Stress Testing AI models the migration cascade — how many accounts migrate between stages at the stressed default rate, what the incremental provisioning requirement is, and whether the resulting Tier 1 capital level remains above the regulatory CET1 ratio floor of 15% for banks.

The rate shock output: three scenarios across the institution's full portfolio

Rate Shock Analysis — California bank · Portfolio $2,840 Crore · Nov 14, 2025
Instantaneous shock model · 3 scenarios · All 4 channels · Computed in 4 minutes 22 seconds
Scenario 1 — Mild (100bps shock)
Shock magnitude+100 bps
New CoF (weighted avg)8.90% (was 7.90%)
NIM compression−48 bps → NIM: 2.12%
Borrowers at DTI >45%+2,840 borrowers
Additional provisions$34 million
CET1 ratio impact−1.2pp → 17.8% (above 15% floor)
Portfolio NPL / charge-off (stressed)3.8% (from 2.9%)
Verdict
Resilient
Capital buffer remains
2.8pp above floor
Scenario 2 — Severe (200bps shock) — Fed / OCC benchmark
Shock magnitude+200 bps
New CoF (weighted avg)9.90% (was 7.90%)
NIM compression−112 bps → NIM: 1.48%
Borrowers at DTI >45%+7,240 borrowers
Additional provisions$96 million
CET1 ratio impact−3.4pp → 15.6% (above 15% floor, barely)
Portfolio NPL / charge-off (stressed)5.4% (from 2.9%)
Verdict
Vulnerable
CET1 ratio: 0.6pp above floor
Management action required
Scenario 3 — Extreme (300bps shock)
Shock magnitude+300 bps
New CoF (weighted avg)10.90% (was 7.90%)
NIM compression−192 bps → NIM: 0.68%
Borrowers at DTI >45%+14,600 borrowers
Additional provisions$188 million
CET1 ratio impact−6.8pp → 12.2% (BELOW 15% floor)
Portfolio NPL / charge-off (stressed)8.1% (from 2.9%)
Verdict
Capital breach
CET1 ratio falls below regulatory floor
Immediate capital plan required
Segment vulnerability — 200bps scenario (Scenario 2) · loan migration to charge-off by product
Home loan (salaried)
NPL / charge-off: 1.8% → 2.6% (mild migration)
Low risk
SME / small business business loan
NPL / charge-off: 3.8% → 6.2% (significant migration)
High exposure
Personal loan (SE)
NPL / charge-off: 5.2% → 9.8% (high migration)
Critical
LAP
NPL / charge-off: 3.1% → 5.4% · Collateral buffer helps
Moderate
Gold loan
NPL / charge-off: 0.8% → 1.1% · Secured, low rate sensitivity
Resilient
● 200bps scenario: CET1 ratio 15.6% — 0.6pp above regulatory floor · SME / small business and SE personal loan segments most vulnerable · Management action required

What the 200bps result means — and what the management action is

The Scenario 2 result — CET1 ratio at 15.6%, just 60 basis points above the regulatory floor — is not a passing result. It is a "barely passing" result that tells the Board and the MD that the institution has very little CET1 ratio cushion against a rate shock that is within the historical range of Fed / OCC policy cycles. A 200bps shock is not an extreme scenario — it is recent experience. A CET1 ratio that falls to 60 basis points above the regulatory floor under recent experience is a signal that capital adequacy requires attention before the next rate cycle.

The management action the Stress Testing AI recommends in this scenario: increase the fixed-rate proportion of the funding book (locking in a longer-dated cost of funds), reduce the floating-rate personal loan (SE) exposure which shows the steepest NPL / charge-off migration at 200bps, and model the additional capital required to maintain a 1.5pp CET1 ratio buffer above the regulatory floor under a 200bps shock — a buffer that the current book does not provide. These recommendations go to the ALCO with the scenario data that generates them, not as opinions but as model outputs.

4Transmission channels modelled — cost of funds, asset quality, liquidity, and capital adequacy
15.6%CET1 ratio at 200bps — just 60bps above the 15% regulatory floor · Management action recommended
9.8%SE personal loan NPL / charge-off at 200bps — the most rate-sensitive segment · Portfolio concentration signal
4m22sFull three-scenario rate shock model computed — from portfolio data to ALCO-ready output

The stress test that barely passes is not a passing stress test

A CET1 ratio of 15.6% against a 15.0% regulatory floor in a 200bps stress scenario does not mean the institution has passed its stress test. It means the institution has 60 basis points of CET1 ratio buffer against a scenario that has occurred within recent Fed / OCC rate cycles. A second modest shock, or a portfolio-quality deterioration unrelated to rates, could push CET1 ratio below the regulatory floor with no additional rate shock required. The Stress Testing Agent AI does not present a barely-passing result as satisfactory — it identifies the result as a warning signal and generates the specific management actions that would restore an adequate buffer. A stress test that only tells you whether you pass or fail is not a stress test — it is a compliance exercise. A stress test that tells you what to do about it is a risk management tool.

← Back to Stress Testing Agent AI