Credit score, PD, LGD, and EAD generation

Plain language meaning

Credit score, PD, LGD and EAD generation turns model-ready lending features into borrower risk, loss severity and exposure estimates that support underwriting, affordability, pricing, limit setting, provisioning and capital analysis.

This topic is specifically about banking credit risk. It should not drift into transaction routing or generic customer analytics.

In a real bank, this is not a loose technology idea. It is a controlled operating step where customer facts, banking policy, model behaviour, human authority, legal obligations and retained evidence must line up. AI and ML can improve speed, consistency and detection quality, but the bank must still prove why the process was fair, explainable, secure, monitored and fit for purpose.

Where it sits in the banking AI journey

This card belongs to The Complete Pipeline. The working flow is Feature input, Credit score, PD estimate, LGD and EAD, and Decision evidence.

Read the flow as a banking control journey. Each stage needs a source system, a decision purpose, a failure mode, a control owner, a fallback path, a customer-impact view and retained evidence. Without those elements, the bank may have automation, but it does not yet have a bank-grade AI process.

Banking data and evidence

The important data points are application score, risk grade, probability of default, collateral value, recovery assumption, drawn balance, undrawn limit, and expected exposure. These items matter because they influence lending eligibility, affordability, fraud risk, compliance treatment, operational queueing, regulatory reporting, customer explanation and audit traceability.

The evidence pack should include score output, PD band, LGD estimate, EAD estimate, policy rule, reason code, and underwriter note. A strong bank can replay the case from source data to feature values, model output, control result, human review, final outcome and monitoring result. A weak bank only knows that a system produced an answer.

Controls that make AI adoption safe

The core controls are approved model version, calibration test, risk-grade mapping, affordability overlay, policy threshold, reason-code validation, and override control. These controls make the topic bank-grade because they tie technical output to approved policy, legal obligations, model governance, operational resilience and management accountability.

AI can help compare records, detect anomalies, retrieve policy, summarise case evidence, prioritise work, highlight weak signals and improve investigator consistency. It should not invent missing facts, ignore failed checks, bypass authority, hide uncertainty, decide material customer outcomes without approval or create explanations that cannot be tied back to approved sources.

Regulatory and governance lens

For banking use cases, model risk, fair lending, adverse-action explanation, credit-risk governance, data lineage, operational resilience, AML/CFT risk-based controls, sanctions compliance, fraud information sharing and auditability can meet in the same workflow. The practical design must therefore be narrower and more disciplined than a generic AI design.

The practical test is simple: if a reviewer asks why the bank used the data, why the model output was trusted, why the customer received that action, why an alert was cleared, why an exception was approved, or why a regulatory record was prepared, the evidence must already exist.

Diagram walkthrough

Read the diagram from left to right as Feature input, Credit score, PD estimate, LGD and EAD, and Decision evidence. The diagram is a control map, not decoration. It shows the minimum route by which data, AI or ML output, human action and audit evidence should connect.

Use it as a 30-minute study method. For each box, ask what system produces the data, what can go wrong, what control detects the weakness, who reviews the case, what customer or regulatory impact could arise and what record proves closure.

Most important mistake to avoid

The common failure is presenting one credit score as the full decision while the bank actually needs separate evidence for default likelihood, loss severity, exposure amount, affordability and policy treatment.

The correction is to slow down the thinking, not necessarily the process. A well-designed banking AI process can be fast, but every fast step must still leave behind source lineage, control evidence, decision reason, human accountability, monitoring data and issue ownership.

Source anchors for accurate study

Federal Reserve SR 26-2, dated 17 April 2026, supersedes SR 11-7 and SR 21-8 for traditional model risk management and clarifies that generative and agentic AI need governance through broader risk-management controls.

NIST AI RMF 1.0 uses Govern, Map, Measure and Manage functions for AI risk management, and NIST AI 600-1 adds generative-AI risk actions for content provenance, hallucination, data protection, cybersecurity and human oversight.

U.S. Regulation B, 12 CFR 1002.9, requires specific principal reasons for adverse action in covered credit decisions, including when a creditor uses an AI model. CFPB Circular 2022-03 was withdrawn on 12 May 2025; do not cite it as current guidance. Primary sources: https://www.consumerfinance.gov/rules-policy/regulations/1002/9 and https://www.consumerfinance.gov/compliance/guidance/withdrawn-guidance/.

The EU AI Act treats AI systems used to evaluate creditworthiness or establish credit scores for natural persons as high-risk, except certain fraud detection and prudential capital contexts.

The Basel Framework IRB standards require banks to estimate and validate PD, LGD and EAD using relevant data, meaningful risk differentiation and ongoing governance.

FFIEC BSA/AML examination guidance expects suspicious activity monitoring systems to be risk-based, explainable by management, periodically reviewed and independently validated where appropriate.

Federal Reserve SR 26-3 and FinCEN's 12 June 2026 Section 314(b) materials clarify fraud-related information sharing under the USA PATRIOT Act safe-harbor framework for participating financial institutions.

OFAC's Framework for Compliance Commitments describes sanctions compliance programme components including management commitment, risk assessment, internal controls, testing and auditing, and training.

Keep the four outputs in their proper roles

A score ranks or groups applicants under a particular model design; it is not automatically a calibrated probability of default. PD needs an event definition and horizon. LGD describes loss severity conditional on default under defined recovery assumptions. EAD estimates exposure at the time of default, which can differ from today's drawn balance when a facility has an undrawn component. For a fictional revolving limit of 10,000 with 4,000 drawn, a bank must not label 4,000 as the future EAD without considering the approved conversion method and behavior before default. Multiplying illustrative values can teach expected loss, but it does not determine the accounting allowance or regulatory capital by itself.

The output record should carry model and calibration versions, observation date, product, horizon, source feature snapshot, limitations and the action that actually followed. An affordability failure can stop approval even when the score is favorable. A manual referral needs its own reason and authority. Analysts should test a missing collateral value, an undrawn limit change, an applicant with no bureau history and a model-service timeout. The customer explanation must trace the actual decision factors and policy, rather than presenting a PD number as the sole reason for an adverse action.

Banking practice note: customer purpose

For credit score, pd, lgd, and ead generation, customer purpose is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from application score to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

AI can assist by comparing records, detecting unusual patterns, retrieving approved policy, summarising weak evidence, prioritising exceptions and preparing review notes. The bank should not allow a generated explanation, a confident score or a convenient dashboard to replace validation, consent, human judgement, customer communication, regulatory judgment or issue closure.

A strong implementation records the source event, data timestamp, consent or lawful basis, model or prompt version, feature values, score or generated output, threshold, reason code, user action, exception status, monitoring result, owner review and final outcome. That record lets risk, compliance, audit, technology and operations speak from the same facts.

Banking practice note: consent and lawful use

For credit score, pd, lgd, and ead generation, consent and lawful use is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from risk grade to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: source lineage

For credit score, pd, lgd, and ead generation, source lineage is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from probability of default to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: KYC and identity

For credit score, pd, lgd, and ead generation, KYC and identity is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from collateral value to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: account behaviour

For credit score, pd, lgd, and ead generation, account behaviour is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from recovery assumption to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: feature freshness

For credit score, pd, lgd, and ead generation, feature freshness is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from drawn balance to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: point-in-time correctness

For credit score, pd, lgd, and ead generation, point-in-time correctness is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from undrawn limit to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: model version

For credit score, pd, lgd, and ead generation, model version is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from expected exposure to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: decision threshold

For credit score, pd, lgd, and ead generation, decision threshold is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from application score to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: reason code

For credit score, pd, lgd, and ead generation, reason code is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from risk grade to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: human review

For credit score, pd, lgd, and ead generation, human review is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from probability of default to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: fraud control

For credit score, pd, lgd, and ead generation, fraud control is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from collateral value to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: sanctions control

For credit score, pd, lgd, and ead generation, sanctions control is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from recovery assumption to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: AML control

For credit score, pd, lgd, and ead generation, AML control is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from drawn balance to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: fair lending

For credit score, pd, lgd, and ead generation, fair lending is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from undrawn limit to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: regulatory reporting

For credit score, pd, lgd, and ead generation, regulatory reporting is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from expected exposure to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: operational exception

For credit score, pd, lgd, and ead generation, operational exception is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from application score to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: customer harm

For credit score, pd, lgd, and ead generation, customer harm is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from risk grade to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: audit trail

For credit score, pd, lgd, and ead generation, audit trail is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from probability of default to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: data quality

For credit score, pd, lgd, and ead generation, data quality is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from collateral value to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: privacy minimisation

For credit score, pd, lgd, and ead generation, privacy minimisation is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from recovery assumption to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: committee reporting

For credit score, pd, lgd, and ead generation, committee reporting is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from drawn balance to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: reconciliation

For credit score, pd, lgd, and ead generation, reconciliation is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from undrawn limit to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: exception ownership

For credit score, pd, lgd, and ead generation, exception ownership is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from expected exposure to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: monitoring cadence

For credit score, pd, lgd, and ead generation, monitoring cadence is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from application score to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: closure evidence

For credit score, pd, lgd, and ead generation, closure evidence is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from risk grade to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: policy retrieval

For credit score, pd, lgd, and ead generation, policy retrieval is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from probability of default to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: investigator feedback

For credit score, pd, lgd, and ead generation, investigator feedback is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from collateral value to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: model drift

For credit score, pd, lgd, and ead generation, model drift is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from recovery assumption to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: control attestation

For credit score, pd, lgd, and ead generation, control attestation is not a side detail. It decides whether the bank can connect the AI or ML output to a real banking purpose, a real customer or regulatory outcome and a real accountable owner. Study the topic as a banking process first and a model process second.

Trace one item from drawn balance to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: customer purpose

Trace one item from undrawn limit to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: consent and lawful use

Trace one item from expected exposure to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: source lineage

Trace one item from application score to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: KYC and identity

Trace one item from risk grade to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: account behaviour

Trace one item from probability of default to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: feature freshness

Trace one item from collateral value to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: point-in-time correctness

Trace one item from recovery assumption to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: model version

Trace one item from drawn balance to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: decision threshold

Trace one item from undrawn limit to EAD estimate. Then ask which control from affordability overlay proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: reason code

Trace one item from expected exposure to policy rule. Then ask which control from policy threshold proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: human review

Trace one item from application score to reason code. Then ask which control from reason-code validation proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: fraud control

Trace one item from risk grade to underwriter note. Then ask which control from override control proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: sanctions control

Trace one item from probability of default to score output. Then ask which control from approved model version proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: AML control

Trace one item from collateral value to PD band. Then ask which control from calibration test proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Banking practice note: fair lending

Trace one item from recovery assumption to LGD estimate. Then ask which control from risk-grade mapping proves the item was complete, current, authorised, relevant and fit for use. If that trace cannot be shown without manual guessing, the process is not yet bank-grade.

Distinct outputs and a policy action

An application model might produce a score and calibrated PD for a stated default horizon. An LGD estimate depends on recovery assumptions and product security; EAD depends on balance and possible future draw. Present their definitions, cohorts and versions before combining them into an expected-loss illustration. A score band is not an approval decision and does not establish affordability.

For an illustrative secured loan and revolving line with the same PD, compare exposure and recoverability. A change in requested limit affects EAD but need not change the borrower's PD. Check that inputs were available at application time and that component models refer to compatible horizons and products. Record policy threshold, affordability check, human referral and final booking separately. Validate on mature outcomes, noting that declined applicants lack repayment labels on the proposed loan.