Better Mortgage partnered with Sei to automate quality assurance on its loan officer communications. Sei ingested hundreds of the lender’s SOP documents — including the edge cases that previously lived only in the compliance team’s heads — and now scores every loan officer call each month against those SOPs and applicable regulations.
Overview
Loan officer conversations are one of the highest-risk surfaces in mortgage origination. UDAAP, TILA, RESPA, and Fair Housing requirements all show up live, in the moment, on every borrower call. The lender’s compliance team had a clear standard for what good looked like — but no way to verify it across the full call volume.
Challenge: Manual QA Could Not Keep Up With Loan Officer Volume
Before Sei, the compliance team manually reviewed a small sample of loan officer calls each day. The math did not work:
- 4 calls a day was the realistic ceiling for human reviewers — anything more meant rushed reviews and missed signals.
- A standing backlog of 1,000+ unaudited calls — meaning most loan officer interactions were never reviewed at all.
- Tribal SOPs locked in compliance reviewers’ heads — hundreds of edge cases and exceptions that were never written down, so newer reviewers caught fewer issues.
- Sampling-based QA missed pattern-level issues — when only a fraction of calls are reviewed, recurring agent-level or scripting-level violations stay hidden.
What Sei did: codify every SOP and score every call
Sei sat down with the compliance team and ingested every SOP document, every regulator-driven guideline, and the long tail of edge cases that had only ever existed as institutional memory. The result is a single rulebook that runs against every loan officer call.
Capabilities deployed:
- Hundreds of SOP documents ingested with precision — every nuance and exception codified, version-controlled, and tested against the full monthly call volume.
- Guideline adherence scoring — Sei catches loan officers deviating from approved scripts, disclosure language, or SOP-required steps, with the violating moment of the call cited.
- Customer complaint detection — early signals of frustration, dispute language, or potential UDAAP exposure are flagged before they escalate into formal complaints.
- Auto-generated agent scorecards — every loan officer gets a per-call and rolling scorecard with the exact moments and rules driving the score, so coaching becomes specific instead of generic.
- SOP-gap insights for compliance and product — when many officers fail the same rule the same way, Sei surfaces it as a likely SOP or training gap, not just an individual issue.
Outcome: No More Sampling and a Cleared Backlog
The numbers from the engagement:
- Call auditing went from 4 calls per day to 1,000+ calls audited every month, with a month of volume clearing in under an hour.
- No more sampling on loan officer calls — audits run across the full month of calls instead of whatever a reviewer could reach.
- 4× analyst productivity on the compliance review team.
- The standing 1,000+ call backlog dropped to zero and stays there, even as monthly call volume grows.
- Compliance flags reduced over time — issues are surfaced and coached against early, before they recur or escalate.
- Detailed insight reports gave compliance and product teams visibility into systemic SOP and training gaps that were previously invisible.
Looking Ahead
With loan officer call QA fully automated, the lender is extending Sei into adjacent surfaces — chat transcripts, email correspondence, and post-close servicing calls. The same pattern holds: codify the rulebook once, score every interaction, and let compliance and coaching focus on the exceptions and the patterns instead of the sampling.