Case Study
UK lender reduces complaint escalation
with real-time vulnerability detection.
How a mid-tier UK lender deployed EchoDepth to detect vulnerability during affordability interviews — reducing FCA complaint escalations and creating documented Consumer Duty compliance.
Does AI vulnerability detection find cases that trained staff miss?
In a twelve-week deployment across a UK lender’s video affordability interviews, approximately 60% of the vulnerability flags raised by EchoDepth had not been identified by the interviewing staff. These were not failures of diligence. They were interactions where the customer presented as composed, answered questions appropriately and gave no verbal indication of difficulty, while the involuntary signals — facial Action Units, voice prosody, linguistic hedging — diverged from what was being said. A trained interviewer reads what a customer chooses to show; the measurement reads what they do not choose. This is the gap that sampling-based quality assurance cannot close, because the interactions most likely to become complaints are the ones that appear unremarkable at the time and are therefore never selected for review. Figures are from an anonymised deployment and have been adjusted to protect client confidentiality.
Note: This case study is anonymised at the client's request. The scenario, findings and outcomes are representative of EchoDepth's deployment in a UK lending environment. Specific figures have been adjusted to protect confidentiality while preserving the relative scale of impact.
The challenge
A mid-tier UK lender processing approximately 40,000 personal loan applications per year identified a recurring pattern: a proportion of declined applicants were filing complaints that escalated to the Financial Ombudsman Service (FOS). Internal review showed that in many of these cases, the applicant had displayed signs of emotional distress during the affordability interview — but these signals had not been identified or documented at the time of the decision.
Under FCA Consumer Duty, the lender needed to demonstrate that vulnerability was assessed before consequential decisions were communicated. Their existing process — post-call quality review of a random 5% sample — was not sufficient. The gap was not process. It was detection.
The deployment
EchoDepth was deployed as a 12-week proof of concept across the lender's video affordability interview workflow. The integration:
- EchoDepth analysed video and voice streams during live affordability interviews via WebSocket
- VAD scores were computed per frame with baseline calibration from the session opening
- Alert thresholds were configured for the lending context: arousal > 0.80, valence < 0.25, dominance < 0.30
- When thresholds were breached, staff received a real-time flag recommending pause, re-routing or additional support
- Full session reports were generated for every interview — creating the Consumer Duty audit trail
What was found
The most significant finding was that approximately 60% of vulnerability flags identified by EchoDepth had not been detected by the interviewing staff. These were not cases of negligence — they were cases where the customer presented as calm and cooperative while physiological markers indicated significant distress. This is precisely the gap that text and keyword-based systems cannot close.
The outcome
Following the PoC, the lender moved to a full deployment across their affordability interview workflow. Key outcomes:
- Consumer Duty documentation — every interview now generates a timestamped audit trail documenting that vulnerability was assessed before the lending decision
- Staff confidence — interviewers reported that the real-time flag gave them "permission" to pause and redirect conversations they instinctively felt were difficult but had no data to support
- Compliance team efficiency — quality review shifted from random sampling to targeted review of flagged sessions, reducing review volume while increasing coverage of high-risk interactions
- Reduced escalation trajectory — early indicators showed a measurable reduction in complaints reaching the FOS from the affordability workflow
Interested in a proof of concept? EchoDepth PoCs are structured around your data, your workflows, and your compliance requirements. From £50,000, including full ROI analysis. Book a discovery call →
What results have EchoDepth financial services deployments achieved?
EchoDepth financial services deployments have demonstrated: 3.1x ROI on compliance technology investment, 40–60% reduction in escalation costs from earlier vulnerability identification, coverage of 100% of interactions versus the 2–3% typical of manual QA, and documented FCA audit trail evidence for Consumer Duty proactive identification requirements.