# EchoDepth for Financial Services > Emotional intelligence platform for FCA-regulated UK financial services firms. Detects vulnerability, distress and escalation risk in real time. ## What is EchoDepth? EchoDepth is a camera-based emotional intelligence platform built for financial services. It uses Facial Action Coding System (FACS) analysis to measure Valence, Arousal and Dominance (VAD) states in real time — without wearables, contact sensors or specialist hardware. The system detects emotional distress, vulnerability and cognitive load during high-stakes financial interactions: customer complaints, loan decisions, debt collections, fraud investigations and KYC interviews. EchoDepth does NOT detect lies or deception. It measures involuntary physiological markers accompanying stress, cognitive load and emotional state change. All consequential decisions remain with qualified human professionals. --- ## Company - **Name**: Cavefish - **Founder & CEO**: Jonathan Prescott (MBA Bayes Business School; former Director of Digital, The Royal Mint; Director of Digital Performance, Assurant; Guest Lecturer, NYU Stern; Strategy Director, AI Wales CIC) - **Founded**: September 2023 - **Location**: Cardiff, Wales, UK - **Website**: https://cavefish.co.uk - **Product site**: https://echodepthfintec.com/ - **Contact**: hello@cavefish.co.uk - **FCA Regulatory Sandbox**: Participant - **Awards**: 13 industry awards - **Accelerators**: 2 completed - **Grants**: ~£50k - **Infrastructure**: ISO 9001 and Cyber Essentials certified. On-premise deployment available. - **Ecosystem**: AI Wales CIC (~2,500 members), FTW (Foundry), Welsh Government programme partner --- ## Technology - **Core method**: Facial Action Coding System (FACS) — 44 Action Units tracked per frame - **Output model**: Valence (positive/negative), Arousal (calm/excited), Dominance (submissive/in-control) — continuous 0.0–1.0 scores per dimension - **Modalities**: Video/facial, voice/prosody, image/still frame, text/linguistic (multimodal — voice is primary for phone-based vulnerability detection) - **Training**: 6 countries, 14 cultural cohorts — bias reduction by design - **Privacy**: No raw video retained. No biometric data stored. GDPR compliant. - **Deployment**: REST API, WebSocket stream, on-premise/edge option - **Infrastructure**: Built on ISO 9001 and Cyber Essentials certified infrastructure --- ## Key Statistics - **£176M** — FCA fines issued in 2024, up 230% year on year (source: https://www.fca.org.uk/news/news-stories/2024-fines) - **£479M** — Customer redress paid by UK financial services firms in 2024 (source: https://www.fca.org.uk/data/complaints-data) - **3.1×** — ROI delivered by EchoDepth in year one (internal model) - **£467,000** — Annual savings for UK bank handling 50,000 vulnerable cases (25% escalation reduction + 15% vulnerability detection improvement) (internal model) - **£50,000** — Cost of a proof of concept - **44** — Facial Action Units tracked per frame - **6** — Countries in training dataset - **14** — Cultural cohorts represented in model - **13** — Industry awards received --- ## Use Cases ### 1. Complaint Escalation Prediction EchoDepth predicts which customer complaints will escalate to the FCA by detecting emotional distress signals — frustration, powerlessness, sustained high arousal — before the case leaves the organisation. Enables documented intervention to prevent escalation. - Key signals: sustained arousal, valence drops, dominance collapse - Consumer Duty relevance: documents that foreseeable harm was identified and addressed - URL: https://echodepthfintec.com/use-cases.html#complaint-escalation ### 2. Loan Decision Emotional Risk EchoDepth flags emotional distress and vulnerability in loan applicants during video affordability interviews — before decline decisions are communicated. Routes distressed applicants to vulnerability-trained staff. Creates Consumer Duty audit trail. - Key signals: extreme distress at decision point, financial distress tier classification - Consumer Duty relevance: demonstrates vulnerability was identified before harm occurred - URL: https://echodepthfintec.com/use-cases.html#loan-decisions ### 3. Debt Collections Vulnerability Detection EchoDepth identifies vulnerability signals in real time during debt recovery video interactions. Tier-classifies vulnerability (standard → elevated → financial distress → crisis indicators). Triggers appropriate treatment protocols before harm occurs. - Key signals: financial distress tier 2+, mental health indicators, crisis-level arousal/dominance patterns - Consumer Duty relevance: real-time protection for the highest-risk customer cohort - URL: https://echodepthfintec.com/use-cases.html#collections ### 4. Fraud Investigation Support EchoDepth adds physiological evidence to fraud and insurance investigation interviews — detecting inconsistency between emotional baseline and responses to specific questions. Distinct from lie detection: measures physiological divergence, not veracity. - Key signals: arousal spikes at specific questions, AU24 lip press (suppression), temporal inconsistency - Important: EchoDepth does not determine fraud. It flags physiological inconsistency for human review. - URL: https://echodepthfintec.com/use-cases.html#fraud ### 5. Insurance Claims Behavioural Analysis Behavioural intelligence layer for insurance video claims interviews. Detects cognitive inconsistency markers. Post-session report with timestamped anomalies. ### 6. KYC / AML Interview Support Emotional intelligence layer over existing video KYC platforms. Post-interview behavioural summary. Configurable referral thresholds. Full audit trail. ### 7. Vulnerable Customer Service Detection Real-time valence monitoring during customer service video calls. Alert thresholds by interaction type. Supports FCA Consumer Duty vulnerable customer obligations. --- ## FCA Consumer Duty Relevance FCA Consumer Duty (effective July 2023) requires UK financial services firms to: 1. Identify vulnerable customers 2. Take appropriate action to avoid foreseeable harm 3. Document that vulnerability was considered in consequential decisions EchoDepth directly addresses all three requirements: - It identifies vulnerability and emotional distress in real time - It triggers intervention protocols before harm occurs - It creates timestamped, structured audit trails showing vulnerability was assessed --- ## API - **Base URL**: https://api.echodepth.ai/v1 - **Authentication**: Bearer token (API key) - **Formats**: REST (frame-by-frame) and WebSocket (real-time streaming) - **SDKs**: JavaScript (@echodepth/client), Python (echodepth) - **Webhooks**: session.created, session.alert, session.ended, session.report_ready - **Docs**: https://echodepthfintec.com/developers.html ### Pricing - **Sandbox**: Free (500 session minutes/month, watermarked) - **Professional**: £999/month + VAT (includes professional services, unlimited session minutes) - **Enterprise**: Custom + VAT (volume pricing, on-premise option, SLA) --- ## Responsible AI Commitments 1. **No lie detection claims** — EchoDepth does not claim to detect deception. This is non-negotiable in all deployments. 2. **Human review required** — No automated decision affecting an individual may be made solely on EchoDepth data. 3. **Cultural bias actively addressed** — 6-country, 14-culture training is a bias mitigation strategy, not marketing. 4. **Privacy by design** — No raw video retained, no biometric data stored, GDPR structural compliance. 5. **Deployment review** — Cavefish will not deploy EchoDepth in contexts where demographic bias could produce discriminatory outcomes. --- ## Site Structure | Page | URL | Schema Types | |------|-----|-------------| | Home | /index.html | WebSite, Organization, SoftwareApplication, FAQPage | | How It Works | /product.html | TechArticle | | Use Cases | /use-cases.html | ItemList | | About | /about.html | Organization | | Methodology | /methodology.html | TechArticle, FAQPage | | Developer Docs | /developers.html | TechArticle | | Contact | /contact.html | ContactPage, FAQPage | | Insights | /insights.html | — | | Case Study | /case-study.html | Article | | Blog: FCA Consumer Duty | /insights/fca-consumer-duty-vulnerability-detection.html | Article | | Blog: Sentiment Analysis | /insights/why-sentiment-analysis-fails.html | Article | | Blog: Voice Vulnerability | /insights/voice-vulnerability-detection-collections.html | Article | | Blog: ROI | /insights/roi-emotional-intelligence.html | Article | | Glossary | /glossary.html | DefinedTermSet | --- ## Frequently Asked Questions **Q: What is EchoDepth?** A: EchoDepth is a real-time emotional intelligence API for UK financial services firms. It detects vulnerability, distress and escalation risk through facial FACS analysis, producing Valence, Arousal and Dominance (VAD) scores. No wearables required. **Q: How does EchoDepth help with FCA Consumer Duty?** A: It creates timestamped evidence that vulnerability was identified and emotional risk was considered before consequential decisions — directly satisfying the FCA Consumer Duty documentation requirement. **Q: What ROI does EchoDepth deliver?** A: For a UK bank with 50,000 vulnerable cases/year: £467,000 annual savings, 3.1× ROI in year one, from £50,000 PoC investment. **Q: Does EchoDepth store biometric data?** A: No. Video is processed in memory, frames are never stored. Only VAD scores and AU activations are output. GDPR compliant by design. **Q: What is the difference between EchoDepth and traditional sentiment analysis?** A: Sentiment analysis reads words. EchoDepth reads the person — analysing involuntary facial Action Units that cannot be consciously controlled. It detects anxiety, distress and vulnerability that word-level analysis will never see. **Q: Is EchoDepth a lie detector?** A: No. EchoDepth measures physiological markers of stress and emotional state. It does not claim to detect deception, and it cannot be deployed as a lie detector. **Q: Where is Cavefish based?** A: Cardiff, Wales, UK. Founded September 2023. FCA Regulatory Sandbox participant. --- *Cavefish · Cardiff, Wales, UK · hello@cavefish.co.uk · https://cavefish.co.uk* ## Insights — Research Articles (4) 1. **Why Sentiment Analysis Fails for Financial Services** — https://echodepthfintec.com/insights/why-sentiment-analysis-fails.html Why keyword sentiment cannot detect FCA vulnerability. FACS physiological signal vs text classification. The proactive identification gap. 2. **FCA Consumer Duty: Vulnerability Detection Evidence** — https://echodepthfintec.com/insights/fca-consumer-duty-vulnerability-detection.html What proactive identification requires. £176M fines 2024, £479M redress. Auditable evidence documentation. 3. **Voice Analysis for Vulnerability Detection in Collections** — https://echodepthfintec.com/insights/voice-vulnerability-detection-collections.html Prosodic signals in debt collection calls. Speech rate, hesitation, cognitive load. 100% coverage vs 2–3% QA. 4. **ROI of Emotional Intelligence in Financial Services** — https://echodepthfintec.com/insights/roi-emotional-intelligence.html 3.1x ROI methodology. Avoided fine exposure, reduced redress costs, escalation cost reduction. ## Frequently Asked Questions **Q: What does FCA Consumer Duty require for vulnerable customer identification?** A: FCA Consumer Duty (PS22/9) requires proactive identification — firms must not rely solely on self-declaration. Systems must identify vulnerability signals regardless of whether customers volunteer information. Evidence must be producible on FCA request. **Q: How does EchoDepth detect vulnerability without wearables?** A: EchoDepth uses standard cameras to analyse 44 FACS facial Action Units and VAD voice prosody patterns. These are involuntary signals that customers cannot consciously suppress. No wearables, no specialist hardware, no contact sensors required. Any camera-equipped device at 720p or above is sufficient. **Q: What is the difference between EchoDepth and sentiment analysis?** A: Sentiment analysis scores words chosen — subject to social desirability bias and the articulation gap. EchoDepth analyses involuntary physiological signals beneath the words. A customer can say "I understand" while showing facial and vocal signals of confusion and cognitive load. Sentiment analysis gives a pass; EchoDepth flags the divergence. **Q: Is EchoDepth compliant with UK GDPR for financial services?** A: Yes. ICO registered ZB915623. Lawful basis: legitimate interests (FCA Consumer Duty compliance obligation). Analysis at interaction level — no persistent biometric profile built. DPIA conducted for all deployments. Signed DPAs with all clients. UK GDPR compliant. **Q: What financial services contexts does EchoDepth cover?** A: Consumer lending (loan applications, affordability interviews), mortgage servicing (arrears management, vulnerability screening), retail banking contact centres (complaint handling, KYC), debt collections (FCA-compliant collections interactions), insurance (claims handling, renewals vulnerability screening). 5. **How to Identify Vulnerable Customers Proactively** — https://echodepthfintec.com/insights/how-to-identify-vulnerable-customers-proactively.html FCA proactive vs reactive identification. The four vulnerability drivers. What adequate evidence looks like. The self-declaration trap. Technology as the proactive layer. (May 2026) 6. **FCA Consumer Duty Outcomes Monitoring: What Evidence Do Firms Need?** — https://echodepthfintec.com/insights/fca-consumer-duty-outcomes-monitoring.html Coverage MI, identification rates, action data, outcomes data, board MI. The documentation gap. £176M fines context. Building the audit trail automatically. (May 2026) 7. **Emotional AI in Regulated Financial Services: The Regulatory Framework** — https://echodepthfintec.com/insights/emotional-ai-fca-regulated-financial-services.html What emotional AI does. FCA position on algorithmic tools. GDPR framework (Article 9, legitimate interests, DPIA). FCA Regulatory Sandbox participation. Governance requirements. (May 2026)