Credit Exception Agent AI

Invoked when: an exception flag is raised in the LOS during credit underwriting — triage is completed within 30 minutes of the flag

• –Categorises every exception into the approved exception taxonomy — the defined list of exception categories that the credit policy recognises: bureau score exceptions (applicant falls below the minimum score threshold but meets other criteria), income documentation exceptions (standard income proof is unavailable but alternative documentation is available), LTV exceptions (loan-to-value ratio exceeds the policy maximum for the product), tenure exceptions (requested tenure exceeds the policy maximum), and sector-specific exceptions (applicant is in a sector that requires committee review regardless of other metrics). Each exception is tagged to exactly one primary category; where an application triggers multiple exceptions, each is documented separately.
• –Assesses the exception against the approved exception thresholds — the range within which the credit policy permits exceptions to be considered, versus the absolute limits beyond which no exception is permissible regardless of other compensating factors. A bureau score exception for an applicant at 690 against a 720 minimum is within the permitted exception range; a score of 580 against a 720 minimum is outside it. The triage categorises the exception as within-range (eligible for committee consideration) or out-of-range (ineligible — the application cannot proceed on an exception basis regardless of the approving authority's preference). Out-of-range exceptions are returned to the underwriter with the specific reason, preventing them from entering the approval queue.
• –Identifies compensating factors for within-range exceptions — the elements of the application that partially mitigate the risk introduced by the policy deviation: a long tenure with the employer mitigates an income documentation exception; a significant existing relationship with LendingIQ (clean repayment history on a prior loan) mitigates a bureau score exception. Compensating factors are documented in the exception pack and presented to the approving authority alongside the deviation — an exception request presented without compensating factors forces the approver to assemble this context themselves, slowing the process and increasing the risk of an approval based on incomplete information.

Invoked when: an exception is confirmed within-range — documentation pack is assembled before the exception is routed to the approving authority

• –Assembles the override documentation pack from the LOS and the credit file — the structured record that the approving authority reviews and that forms the audit trail for the exception. The pack contains: the application summary (borrower profile, product requested, amount, tenure), the specific policy deviation (which criterion is breached, by how much, and what the policy requires), the compensating factors (documented with supporting evidence from the credit file), the underwriter's recommendation (approve exception / decline), and the credit score if the exception proceeds — an estimate of the incremental credit risk introduced by granting the exception relative to a standard-policy approval. The pack is formatted to the credit committee's standard template — no reformatting required by the approver.
• –Attaches the relevant supporting documents from the credit file to the pack — the income documents that justify the income documentation exception, the bureau report that shows the score deviation, or the property valuation that establishes the LTV. Every factual claim in the pack is supported by an attached document; unsupported claims in exception packs create inspection findings when the supporting document cannot be produced on request.
• –Generates a one-paragraph exception rationale — a plain-language summary of why the underwriter is requesting the exception, what risk it introduces, and what mitigates that risk. The rationale is the approver's primary read when they first encounter the exception; a clear rationale reduces the back-and-forth between the credit team and the approver that extends the approval turnaround. The rationale is drafted by the agent and reviewed by the underwriter before the pack is submitted — the agent's draft reflects the factual basis; the underwriter adds professional judgment context where relevant.

Invoked when: documentation pack is complete (routing), or monthly analysis cycle runs (pattern analysis)

• –Routes each exception to the correct approval tier based on the exception category and the loan amount — the credit policy defines a matrix of exception categories against amount thresholds that determines whether the exception goes to the credit head (smaller amounts, lower-severity categories), the credit committee (larger amounts or higher-severity categories), or the board credit committee (exceptions above specified thresholds that require board-level oversight). The routing is automatic and deterministic — the agent applies the matrix, inserts the exception as an agenda item in the next scheduled session of the relevant body, and dispatches the documentation pack to the members. Exceptions routed to the wrong approval tier are a common inspection finding; deterministic routing eliminates this risk.
• –Tracks exception approval SLA — the time between the exception being submitted to the approving authority and the approval or decline decision being recorded. The SLA target is 72 hours for credit head approvals and 5 business days for committee approvals. Exceptions that approach the SLA boundary without a decision receive an automatic reminder to the approving authority. SLA breaches are flagged to the CCO; a pattern of SLA breaches in a specific approval tier indicates a capacity or process problem in that tier's review cycle.
• –Produces a monthly exception pattern analysis for the CCO — the frequency of exceptions by category, the approval rate by category and approving authority, the concentration of exceptions in specific product types or borrower segments, and the trend over the prior 3 months. A category where exception volume is rising as a proportion of approvals in that category indicates that the policy threshold may no longer be appropriately calibrated to the market — the exception is becoming the norm, which means either the policy needs updating or the underwriting process is drifting. The pattern analysis is the signal; the policy review is the human decision.