Factually incorrect outputs from artificial intelligence models carry higher career risk in financial institutions than aesthetic or formatting errors