Every underwriting exception, in plain sight
Prisma checks every underwriting decision and exception against your credit policy, flags what deviates or is inconsistent, and keeps the evidence, before a deviation becomes portfolio risk.
Discover the Prisma platformAt a glance
- Department
- Underwriting and credit
- Regulatory context
- Credit underwriting standards and supervisory expectations on tracking policy exceptions, plus fair lending requirements under ECOA and Regulation B (12 CFR 1002)
- What is reviewed
- Underwriting decisions, policy exceptions and credit files, whether prepared by people or with AI assistance
- What Prisma delivers
- Full-population review, escalation of weak and ambiguous decisions, and an audit trail for every decision
Every underwriting decision is a judgment against policy
Underwriting applies a written credit policy to files that are rarely simple: financial statements, collateral, documentation and conditions, with rules that interact and decisions that still call for judgment. Most decisions follow the policy. Some depart from it, and a departure is a policy exception: a decision approved outside the standard criteria, with a recorded reason.
Underwriting decision oversight is the practice of checking those decisions and exceptions against the policy and against the evidence in the file. It asks whether the criteria were applied as written, whether every exception was justified and recorded, and whether similar files led to similar outcomes.
For underwriting and credit, this is where portfolio quality is set. A decision that cannot be explained against policy, or an exception that was never recorded, joins the portfolio long before anyone reviews it.
Why exceptions need watching
Exceptions are a normal part of lending. What matters is whether they are justified, limited and visible. Supervisory guidance expects institutions to track and report their policy exceptions, and the exception rate, the share of decisions that depart from policy, is one of the clearest signals of how underwriting is really being applied.
When exceptions go unrecorded or unjustified, the effects spread. Loans can be priced for a risk that was never assessed, exceptions can cluster in one product or period without anyone seeing it, and similar applicants can receive different outcomes, which is a question that fair lending rules ask about directly.
Where sampling stops
Quality control by sampling reads a small share of decisions, and it can only describe that share. A deviation, an unrecorded exception or an inconsistent outcome outside the sample goes unseen, and a pattern that spans the portfolio never shows up in a handful of files.
Reviewing every decision changes what can be known. Each exception is identified when it is made, and the pattern across products, periods and decisions becomes visible while it can still be corrected.
Reading every decision against your credit policy
Prisma evaluates each underwriting decision against the credit policy and regulatory expectations your institution has set, and returns a result of PASS, FAIL or AMBIGUOUS, together with the reasoning behind it. A PASS means the decision and its documentation meet policy. A FAIL records why they do not. An AMBIGUOUS result means the file does not allow a clear answer, so the case is escalated for a human decision instead of forcing certainty.
The evaluation looks at whether the policy criteria were met, whether any exception is documented and justified, and whether the outcome is consistent with how similar cases were decided.
When a decision falls short, oversight is followed by intervention: the weak decision is escalated for review, and consequences and feedback are applied in time to fix it. Remediation then closes the loop. Findings feed back into how future decisions are made, so the same deviation does not repeat, and the evidence of how each decision was judged is kept.
One standard for people and for AI
Underwriting is increasingly supported by models and AI agents that summarize documents, test criteria and recommend outcomes. The same expectations apply to them: decisions must follow policy, reasons must be specific and consistent, and the institution has to be able to show how each outcome was reached. Prisma applies a single governance standard to work produced by people and by AI, so an AI-assisted decision is read against the same policy and receives the same PASS, FAIL or AMBIGUOUS result as one made by hand. Delivered as an SDK that runs inside your own environment, it adds that oversight to the tools underwriting already uses.
Don’t just take our word for it
Frequently asked questions
Underwriting decision oversight is the review of how underwriting decisions and exceptions are made and documented. It checks that each decision follows your credit policy and is supported by the evidence in the file, across every decision in the portfolio.
A policy exception is a decision that departs from the written underwriting policy, approved with a recorded reason. Exceptions are a normal part of lending, which is why each one needs to be justified, recorded and consistent with how similar cases were decided.
The exception rate is the share of decisions approved outside the standard underwriting criteria. It matters because supervisory guidance expects institutions to track and report exceptions, and because a rising rate, or exceptions concentrated in one product or period, can show that policy is not being applied as intended.
A sample reads only a small share of decisions, so a deviation, an unrecorded exception or an inconsistent outcome outside the sample goes unseen. Prisma reviews every decision against your policy, so exceptions are identified across the whole population.
Fair lending law, including the Equal Credit Opportunity Act and Regulation B (12 CFR 1002), looks at whether similar applicants are treated consistently. When exceptions are granted or refused inconsistently, similar applicants can end up with different outcomes. Reviewing every decision makes those patterns visible.
Regulation B requires a creditor that denies credit to give specific reasons (12 CFR 1002.9), whether or not the decision relied on a model. Prisma reads AI-assisted decisions against the same policy as any other, and records the reasoning behind each result.
When the file does not allow a clear result, Prisma returns AMBIGUOUS and escalates the case for a human decision, so the institution never has to force certainty. The reasoning is recorded with the result, next to PASS and FAIL.
Yes. Prisma evaluates decisions against the policy and regulatory expectations your institution has set, so the standard it applies is yours.
No. Your institution keeps the credit decision. Prisma reviews how decisions and exceptions are made and documented, intervenes when a decision falls short of policy, and keeps the evidence.
Prisma is delivered as an SDK that runs inside your own environment. It can also be delivered through Palqee's managed cloud and interface if you need it.
