Location
Product
Financial institution type
Seacoast Bank set an ambitious goal: automate every nonjudgmental component of its Credit Risk Review process by 2028. By implementing Abrigo's Loan Review Assistant within DiCOM Loan Review, Seacoast significantly reduced the time required to complete complex loan review documentation while maintaining human oversight and regulatory confidence. Read the highlights below, or download the full case study to learn more about Seacoast Bank's experience.
Download the full case studyLike many financial institutions, Seacoast Bank's Credit Risk Reviewers spent a significant portion of every review manually copying information from approval packages into loan review documentation. Borrower summaries, collateral descriptions, underwriting facts, cash flow narratives, and facility information all had to be retyped or copied from multiple source documents before Reviewers could begin the work that required professional judgment.
"We spent the most time copying and pasting or retyping information from our approval documents," said Hannah Primes, Vice President and Senior Commercial Credit Risk Review Officer. "We wanted to automate the work that didn't require judgment so we could spend more time on the work that did."
The team also recognized that adopting generative AI required the right partner and thorough planning. Any solution would need to produce consistent, explainable results that could withstand regulatory scrutiny while preserving the reviewer's responsibility for conclusions.
Seacoast implemented Abrigo's Loan Review Assistant to automate narrative generation directly from loan approval documentation.
The bank began by targeting the highest-volume, most repetitive sections of its line sheets, including:
Rather than asking AI to make credit decisions, the team intentionally limited automation to factual information extracted from source documents.
"The Loan Review Assistant pulls facts, and we come in behind and fill in the blanks with the subjective things we do," Primes explained. "Our Reviewers still determine whether what was done by the line of business meets policy and whether weʼre comfortable with the credit and its risk rating."
Every prompt result is reviewed by a human before becoming part of the final loan review. The team also configured prompts to provide citations linking back to the source documents, allowing Reviewers to quickly verify the origin of each statement. As implementation progressed, Seacoast refined prompts after every review cycle, continuously improving output quality while documenting lessons learned.
After implementing Loan Review Assistant, Seacoast cut the time required to complete complex line sheets nearly in half. Prompt refinements steadily reduced issues while improving consistency across reviews. Seacoast validated every prompt result and is pleased with the toolʼs accuracy.
"We haven't had any inaccurate prompt results,ˮ Primes said. “[Loan Review Assistant] has never made up any facts about the loans. They've always matched the source document directly."
Other key outcomes include:
Time savings: Typical clean linesheets that previously required eight to 10 hours now consistently take four to six hours to complete.
Improved documentation quality: The project reinforced an important lesson: better source documents produce better AI outputs. The team now uses AI feedback to identify inconsistent or incomplete underwriting narratives and shares those findings with lending teams to improve documentation quality upstream.
Human judgment remains central: By limiting AI to nonjudgmental tasks, Reviewers continue to make all credit conclusions, policy assessments, and risk determinations while spending less time on administrative work. Seacoast deliberately keeps AI focused on factual work so Reviewers retain responsibility for their output.
Learn what to ask to understand vendor accountability, explainability, human oversight, and data governance.
