London Market & Lloyd's Advisory · 2026
Strategic AI Playbook

The Discipline
Imperative

Triple Filter™ Agentic AI for London Market Underwriting Sovereignty

A C-suite strategic framework for Lloyd's syndicates and London Market insurers navigating the intersection of AI disruption, broker-driven adverse selection, and the return of the soft cycle — and how to win through systematic underwriting discipline.

🔵 Filter I · Guideline Adherence
🟡 Filter II · Deep Risk Synthesis
🟢 Filter III · Anti-Selection Detection
CH 01
What is Triple Filter?
The strategy explained
CH 02
The Burning Platform
Market forces at inflection
CH 03
The Strategic Gap
Syndicate implications
CH 04
Triple Filter Architecture
The three filters in depth
CH 05
Value at Stake
Quantified outcomes
CH 06
Deployment Approach
Phased programme
CH 07
Risks & Mitigations
What to watch
CH 08
AI at the Tip
Evolving dynamically
CH 09
Outcome Horizon
Short · Mid · Long term

Click a chapter above or use ← → arrow keys · Click any card to explore detail

Confidential · London Market & Lloyd's Advisory · For CUO / CEO / CRO Discussion Only · 2026
Chapter 01 · The Strategy Defined
What is the Triple Filter™ Strategy?
A complete AI underwriting operating system — three interlocked filters that address the three root causes of underwriting loss that human processes alone cannot solve
The Founding Insight
"Every large underwriting loss in the London Market can be traced to one of three failures — a guideline was bent, a hidden hazard was missed, or the wrong broker's risk was accepted. The Triple Filter eliminates all three. Simultaneously."
— The strategic premise underlying the Triple Filter framework
⬡
Triple Filter™
I
Filter I
Absolute Guideline Adherence
"Eliminating the drift that soft cycles always create"
An agentic AI engine enforces 100% compliance with the syndicate's technical underwriting standards — Lloyd's Minimum Standards, SBP commitments, and class-level appetite parameters — on every submission, at every point in the market cycle.
THE PROBLEM IT SOLVES
Human underwriters are rational actors. In a softening market, rational actors make concessions — individually justifiable, collectively catastrophic. AI does not respond to competitive pressure, relationship dynamics, or cycle psychology. It enforces what the CUO intended, without exception.
127 Rules Enforced Zero Cycle Drift Real-Time · Pre-Bind
II
Filter II
Deep Risk Synthesis
"Seeing what brokers choose not to show you"
An AI Synthesis Agent cross-references every submission against 14 external intelligence sources — HSE records, court databases, satellite imagery, credit files, dark web signals, and social data — to surface the hidden hazards that lead to large losses but never appear in the broker's narrative.
THE PROBLEM IT SOLVES
The broker submission is the insured's best case — curated to achieve the most favourable terms. In 62% of large losses, at least one material external signal existed before inception that was not in the submission and not acted upon. Filter II makes this invisible information visible, pre-bind, on every risk.
14 External Sources Unstructured Data Narrative Reconciliation
III
Filter III
Anti-Selection Detection
"Knowing which brokers are sending you their worst"
A predictive analytics engine tracks the quality of business each Lloyd's broker routes to your syndicate versus their broader market placement — identifying adverse selection patterns, calibrating broker quality scores, and systematically prioritising risks from channels with historically superior loss outcomes.
THE PROBLEM IT SOLVES
In the London Market's broker-intermediated model, information asymmetry systematically disadvantages carriers. The best brokers' best risks go to preferred carriers. Lloyd's receives the residual. Filter III quantifies this asymmetry for the first time — turning relationship intuition into data-driven portfolio prioritisation.
CDR-Integrated 47-Feature ML Model Broker Scorecard
How the Three Filters Work Together
I
Selects only compliant risks. Every submission is assessed against 127 rules before human review begins. Non-compliant risks are flagged, referred, or declined — automatically — removing the largest single driver of cycle deterioration.
↓
Compliant risks proceed to Filter II
II
Validates the compliant risks against reality. External intelligence is synthesised against the broker's narrative. Hidden hazards are surfaced. The underwriter receives a reconciliation report before making any decision.
↓
Validated risks assessed for broker quality
III
Prioritises by broker channel quality. The source of every submission is scored against its historical track record. Risks from low-quality broker channels face enhanced scrutiny regardless of individual risk merit — because channel quality is a proven predictor of portfolio outcome.
↓ Composite score → Approve / Refer / Decline + Priority-ranked Next Best Actions
What Makes This Different From Other AI Underwriting Tools
Conventional AI Tools
Single-dimension scoring — typically price or risk quality alone
Triple Filter™
Three interlocked failure mode responses — compliance, intelligence, selection
Static model — trained once, applied until outdated
Continuous learning loop — retrained monthly on closed loss outcomes
Submission data only — relies on what the broker provides
14 external intelligence sources — sees beyond the broker's narrative
Risk-level view only — no broker channel intelligence
Portfolio-level broker quality scoring — CDR-linked, real-time
Black-box outputs — not auditable for SMCR or PMD
Full explainability — every recommendation attributed to specific inputs
Chapter 02 · The Strategic Context
The Burning Platform: Why Now, Why London
Six converging forces are reshaping the London Market's underwriting landscape — and the window to act is narrowing
£3.4B
Lloyd's Underwriting Loss
2017 Peak — Decile 10
↑ Drift-driven deterioration
−6.1%
London Market Rate Change
2024 — Softening Resumes
↓ Cycle turning
68%
London Risks Placed via
Top 5 Broker Groups
→ Concentration risk elevated
$4.1B
Global InsurTech AI Investment
2023–24 — Competitors arming
↑ AI capability gap widening
📉
The Soft Cycle is Returning — Faster Than Expected
London Market composite rates declined for the first time since 2018 in H2 2024, with Property D&F and Casualty XL leading the softening
Lloyd's PMD cycle data shows a 14-point combined ratio deterioration in the last three soft cycles — driven almost entirely by guideline drift at the point of inception
The post-COVID hard market correction has now largely been absorbed, and capacity is returning faster than demand growth
Critical
🎯
Broker Adverse Selection: A Structural, Measurable Problem
In the Lloyd's broker-led market, the largest intermediaries have unambiguous incentive to place the most challenging risks at Lloyd's while retaining quality for tier-one company market capacity
Internal Lloyd's CDR data consistently shows 8–15% higher attritional loss ratios on broker-driven versus direct placements in equivalent risk classes
No systematic, data-driven broker quality scoring exists at most Lloyd's syndicates today — it is managed through relationship intuition alone
Structural Risk
🌍
External Risk Complexity is Accelerating Beyond Human Synthesis
Climate model uncertainty has materially expanded the gap between broker-declared exposure and modelled loss potential — particularly in UK Flood, European NAT CAT, and US Coastal Property
Cyber threat velocity has rendered static application-based cyber underwriting fundamentally inadequate — real-time intelligence from NCSC, CISA, and dark web monitoring is now table stakes
Social inflation in UK Casualty is running at 8–12% per annum above RPI, invisible in historical loss runs but highly material in forward exposure
Growing Rapidly
⚖️
Regulatory & Supervisory Pressure Intensifying at Lloyd's
Lloyd's PMD has significantly strengthened SBP (Syndicate Business Plan) scrutiny following the post-2017 remediation period — with increased focus on underwriting controls and evidenced compliance
PRA SS1/23 requirements on model governance and SMCR accountability create personal liability for senior underwriting managers who cannot evidence systematic controls
DORA (Digital Operational Resilience Act) creates EU-facing obligations for Lloyd's syndicates writing European risks — with AI system governance now in scope
Regulatory Pressure
🤖
AI-Native Competitors are Redefining the Speed-to-Quote Benchmark
Startups including Kaia, Planck, and CyberCube have deployed AI underwriting tools that reduce submission-to-quote time from days to minutes while maintaining or improving loss selection
Munich Re, Swiss Re, and Zurich have invested >$800M collectively in AI underwriting platforms since 2022 — the capability gap with traditional Lloyd's syndicates is growing
Brokers are increasingly routing price-sensitive risks to AI-enabled carriers first, leaving London Market syndicates with the residual adverse selection
Competitive
👥
Underwriting Talent Scarcity and Institutional Knowledge Erosion
An estimated 30–35% of senior Lloyd's underwriting talent will reach retirement age by 2030 — creating an urgent institutional knowledge transfer problem
Junior underwriters entering the market have 40% less mentored experience than equivalent cohorts in 2010, increasing the risk of guideline drift through inexperience rather than intent
AI-encoded underwriting standards represent the most durable solution to knowledge preservation — independent of individual tenure or cycle pressures
Opportunity
Chapter 03 · Syndicate Implications
The Strategic Gap: What This Means for Your Syndicate
Four structural vulnerabilities that the Triple Filter strategy directly addresses — each with measurable loss impact
📐
14 pts
Combined ratio deterioration — last three soft cycles (Lloyd's average)
Guideline Drift is Systemic, Not Individual
Every soft cycle produces the same pattern: underwriters soften requirements incrementally, each individually rational, but collectively catastrophic. The 2017 Lloyd's loss was not caused by bad individual decisions — it was caused by 1,000 small concessions that were invisible until claims arrived. AI enforcement eliminates this entirely by removing the human discretion that creates drift.
Most Urgent
⚖️
8–15%
Higher attritional LR on broker-driven Lloyd's placements vs. direct
Adverse Selection is Measurable but Unmanaged
The London Market's broker-intermediated model creates structural adverse selection that is well understood in principle but almost never quantified in practice. Syndicates lack the systematic data infrastructure to track broker quality across the full portfolio and correlate it with loss outcomes at granular enough resolution to act. The Triple Filter's Anti-Selection engine makes this visible, persistent, and actionable.
Structural
🔎
62%
Of large losses have at least one external signal available pre-inception that was not acted upon
Hidden Hazards Grow Faster Than Disclosed Risks
Research across multiple Lloyd's classes consistently shows that the majority of large losses have at least one materially relevant data signal available before binding — in OSHA records, court databases, satellite imagery, or social intelligence — that simply was not visible to the underwriter at the time of decision. The broker's narrative systematically omits what is inconvenient. AI synthesis removes this asymmetry.
High Impact
⚡
3.2 days
Average Lloyd's submission-to-quote time vs. sub-4hrs for AI-native carriers
Speed Asymmetry Creates Adverse Selection Feedback Loop
When AI-native competitors quote in hours and Lloyd's syndicates quote in days, brokers route their most straightforward, best-quality risks to the fast carriers first. Lloyd's then sees the residual — the complex, the marginal, and the adverse. The Triple Filter accelerates decision quality without sacrificing rigour, enabling syndicates to compete on both speed and selection simultaneously.
Competitive
The CUO's Dilemma — Four Impossible Asks, One Answer
🎯
Maintain underwriting quality
Despite competitive pressure to soften terms during cycle downturn
⚡
Compete on speed
Against AI-native carriers quoting in hours with smaller UW teams
🤝
Manage broker relationships
While systematically declining the risks they most want to place
📈
Grow the portfolio
Without proportionally growing the underwriting headcount and cost base
The Triple Filter resolves all four simultaneously — through agentic AI that is faster, more consistent, and more data-rich than any individual underwriter.
Chapter 04 · The Three Filters
Triple Filter™ Architecture: Three Layers of Underwriting Intelligence
Each filter addresses a distinct failure mode. Together they constitute a complete AI underwriting operating system
I
Absolute Guideline Adherence — Eliminating Drift
Agentic AI enforces 100% compliance with Syndicate UW standards, Lloyd's Minimum Standards and SBP commitments · Unlike human UW, AI does not soften in a soft cycle
127 Rules · Real-Time · Zero Exceptions
01
Lloyd's Minimum Standards Compliance Engine
02
SBP Appetite Gate — Real-Time
03
Market Cycle Drift Guard
Key Innovation
04
Limits, Deductibles & Sublimit Audit
05
PRA/FCA/SMCR Accountability Trail
06
Exclusion & Endorsement Validation
II
Deep Risk Synthesis — Moving Beyond the Broker's Narrative
AI Synthesis Agent cross-references submissions against HSE, UK Land Registry, Companies House, JBA Flood, ICO, NCSC, satellite imagery and social sentiment to surface hidden hazards
14 External Sources · Unstructured Data · Pre-Bind
01
HSE / RIDDOR Violation Intelligence
02
JBA Flood / OS MasterMap Geospatial
03
Companies House & HMCTS Records
04
NCSC / ICO / Dark Web Intelligence
Cyber Line
05
Social Sentiment & ESG Signals
06
Broker Narrative vs. AI Synthesis Reconciliation
Key Output
III
Anti-Selection Detection — Quantifying Broker Selection Bias
Predictive analytics engine tracks quality of business from each Lloyd's registered broker vs. broader market placement, prioritising risks from brokers with historically superior combined ratios
CDR Integrated · Portfolio-Level · Broker Scorecard
01
Broker Quality Scorecard (CDR-Linked)
Core Asset
02
Placement Pattern Bias Detection
03
Coverholder & MGA Delegated Auth Analysis
04
XGBoost Loss Probability Model (47 Features)
05
Risk-Adjusted Pricing Recommendation
06
Broker Channel Prioritisation Matrix
Chapter 05 · The Business Case
Value at Stake: Quantifying the Triple Filter Advantage
Conservative estimates based on Lloyd's CDR benchmarking, PMD data and comparable AI underwriting deployments across global carriers
3–8 pts
Combined Ratio
Improvement (Year 3+)
↑ Primary value driver
12–18%
Loss Ratio Reduction
in affected portfolios
↑ Hidden hazard elimination
73%
Reduction in submission
triage time (hrs → mins)
↑ Capacity freed for complex
6–9%
Capital efficiency gain
(Solvency II Model output)
↑ Via improved risk selection
Value Decomposition by Filter
🔵 Filter I — Guideline Enforcement
2–4 CR pts
Eliminates soft-cycle drift premium leakage. Lloyd's historical data shows ~2 CR points gained per cycle through consistent enforcement. Most impactful during market softening phases.
🟡 Filter II — External Data Synthesis
1–3 CR pts
Pre-bind identification of hidden hazards prevents large loss exposure. Most impactful in Property D&F (satellite), Casualty (HMCTS/HSE), and Cyber (NCSC/dark web). Estimates based on historic pre-bind signal availability.
🟢 Filter III — Anti-Selection Detection
1–2 CR pts
Systematically re-routing capacity towards higher-quality broker channels improves portfolio average. Effect compounds as broker behaviour adapts in response to selection signals over 18–24 months.
⬡ Composite — Triple Filter Combined
3–8 CR pts
Filters are multiplicative, not additive. Quality selection (F3) surfaces better risks; external synthesis (F2) validates them; guideline enforcement (F1) ensures discipline throughout the cycle. Together they constitute a complete underwriting operating system.
Indicative Financial Impact
MetricBaselineWith Triple FilterDelta
Combined Ratio96.2%89–93%−3 to −8 pts
Loss Ratio64.1%54–58%−6 to −10 pts
Expense Ratio32.1%30–31%−1 to −2 pts
Quote Cycle (hrs)76 hrs4–18 hrs−70–90%
UW Capacity (subs/UW)180/yr420–600/yr+2–3×
Top-Tier Broker Mix44%62–70%+18–26 pts
SCR Reduction (est.)—6–9%Capital freed
Indicative estimates based on Lloyd's CDR benchmarking data and comparable AI deployment case studies. Actual outcomes depend on portfolio composition, data maturity, and deployment scope. Independent validation recommended.
Chapter 06 · Programme Approach
Deployment: A Phased, Low-Disruption Programme Architecture
Three phases — Foundation, Intelligence, and Full Agentic Loop — each delivering standalone value and building toward complete underwriting sovereignty
1
Phase 1 · Months 0–4
Foundation: Data, Architecture & Guideline Encoding
CDR data pipeline established. Syndicate UW guidelines encoded into AI rule engine. Broker quality baseline scored. Lloyd's data taxonomy mapped and validated. Shadow mode deployment begins — AI runs parallel to existing UW process.
Shadow Mode Data Pipeline Guideline Encoding
2
Phase 2 · Months 4–10
Intelligence Layer: External Data Synthesis & Broker Scoring
14 external data sources integrated (HSE, JBA, Companies House, HMCTS, NCSC/ICO, Experian, OS MasterMap). Synthesis Agent trained on class-specific loss patterns. Broker anti-selection model calibrated against 36-month CDR history. AI recommendations introduced as mandatory UW input — final decision remains with human UW.
External Data Live Broker Scoring NBA Prompts
3
Phase 3 · Months 10–18
Full Agentic Loop: Autonomous Triage, Guided Bind
AI autonomously triages >70% of submission volume below appetite thresholds (decline / referral) within 4 hours. Complex, above-threshold risks surface to senior UW with full Triple Filter analysis pre-completed. Continuous learning loop active — models retrained monthly on closed loss data. PMD reporting automated.
Agentic Triage Auto-PMD Reporting Continuous Learning
Programme Gantt — Illustrative
WorkstreamM1M2M3M4M5M6M7M8M9M10M11M12M13-18
Data Architecture & CDR Pipeline
Guideline Encoding Engine
Shadow Mode Validation
External Data Integration (14 sources)
Broker Scoring Model Build
NBA Prompt Engine & UW Workflow
Agentic Triage Deployment
PMD Reporting Automation
Continuous Learning Loop Active
★ Phase GatesPhase 1 GoPhase 2 GoPhase 3 GoBAU
Governance
AI Underwriting Oversight Board
CUO-chaired. Monthly review of model performance, guideline updates, broker scoring recalibration. Quarterly PMD reporting input.
Change Management
UW Champion Network
Senior UW champions embedded in each class. AI positioned as analytical co-pilot, not replacement. Decision authority remains with UW.
SMCR / Accountability
Full Audit Trail by Default
Every AI recommendation, human override, and NBA acceptance is timestamped, attributed, and stored for PRA/FCA evidential requirements.
Chapter 07 · Risk Management
Risks & Mitigations: What to Watch — and How to Protect Against It
Nine material risks across Model, Data, Regulatory, Operational, and Strategic dimensions — each with a structured mitigation posture
Model Risk
HIGH
AI Model Overfitting to Historical Loss Patterns
Models trained on past cycles may fail to identify emerging risks in new classes. Climate model shift and novel cyber threats are paradigmatic examples. Mitigation: Monthly retraining cycles, out-of-sample validation, mandatory human UW override for emerging risk classes.
Owner: CRO / Chief Analytics Officer
Data Risk
HIGH
CDR Data Quality & Completeness Gaps
Lloyd's CDR data has known quality gaps — particularly in delegated authority, binder, and older class business. Garbage-in AI outputs destroy underwriter trust rapidly. Mitigation: Pre-deployment data audit, CDR reconciliation protocol, confidence scoring on all AI outputs, fallback to human UW when data below threshold.
Owner: COO / Data Governance Lead
Regulatory Risk
HIGH
PRA / FCA AI Governance Requirements (SS1/23)
PRA SS1/23 requires material model governance for AI systems in insurance underwriting. Failure to evidence model oversight creates SMCR personal liability for CUO/CEO. Mitigation: Model Risk Policy updated pre-deployment. Independent model validation. Full explainability layer. PMD pre-notification recommended.
Owner: CRO / General Counsel
Relationship Risk
MEDIUM
Broker Relationship Disruption from Systematic Declination
Filter III will systematically decline more risks from lower-tier brokers. This creates relationship tension if not managed proactively. Mitigation: Broker engagement programme — share quality scorecards with brokers, position as a collaborative quality improvement tool. Frame as "better partnership, not penalisation."
Owner: Active Underwriter / BD Director
Culture Risk
MEDIUM
Underwriter Resistance and AI Override Culture
If UWs routinely override AI recommendations without consequence, the system loses effectiveness and creates false confidence. Mitigation: Override tracking and review embedded in governance. Override patterns reviewed by CUO monthly. Positive framing — AI as analytical co-pilot, not adversary or performance management tool.
Owner: CUO / HR Director
Vendor Risk
MEDIUM
External Data Provider Dependency & API Reliability
14 external data integrations create significant dependency on third-party API uptime and licensing continuity. Mitigation: Waterfall fallback architecture — AI operates on reduced data if external sources are unavailable, with confidence score adjusted. Contractual SLA requirements on all data providers. No single-source dependency for critical Filter II signals.
Owner: CTO / Procurement
Fairness Risk
MEDIUM
Algorithmic Bias in Broker or Risk Classification
AI models may encode historical biases — systematically disadvantaging certain geographies, sectors, or broker types in ways that are commercially or legally problematic. Mitigation: Bias audit as part of Phase 1 data assessment. Protected characteristic variables excluded. Annual fairness testing. FCA Consumer Duty considerations reviewed for retail-adjacent lines.
Owner: CRO / Chief Compliance Officer
Talent Risk
LOWER
AI Skill Gap in UW and Technology Teams
Neither the UW team nor the technology function may have sufficient AI literacy to deploy and maintain the Triple Filter effectively. Mitigation: Embedded AI capability building programme. Senior data science lead hired as dedicated Triple Filter product owner. Partnership with specialist AI insurance vendor for model build and MLOps infrastructure.
Owner: CEO / CHRO
Programme Risk
LOWER
Scope Creep and Phased Delivery Discipline
AI programmes in insurance frequently expand ambition beyond delivery capacity, stalling at proof-of-concept without reaching production impact. Mitigation: Strict phase gate governance. Minimum viable deployment focus — three classes, one pilot syndicate. Standalone value required at each phase gate before continuation is approved.
Owner: COO / Programme Director
Chapter 08 · Dynamic Market Adaptation
AI at the Tip of the Spear: Evolving in a Dynamic Market
The Triple Filter is not a static deployment — it is a living underwriting intelligence system that improves with every submission, every loss, and every market cycle
The AI Evolution Roadmap
1
Continuous Learning — Monthly Loss Feedback Loop
Closed loss data automatically re-trains all three Filter models monthly, ensuring the system self-improves against actual outcomes
Live — Phase 3
2
Market Cycle Intelligence — Automatic Guideline Recalibration
Real-time Lloyd's PMD rate data and market position metrics automatically flag when Filter I thresholds require CUO-approved recalibration for market responsiveness
Phase 3+
3
Emerging Risk Class Extension — Climate & Parametric
The Filter II Synthesis Agent extends to climate transition risk signals, physical risk geospatial layers, and parametric trigger calibration for new class expansion
Yr 2 Horizon
4
Cross-Syndicate Benchmarking via Lloyd's Data Alliance
Participation in a Lloyd's-level AI data consortium enables cross-syndicate loss pattern benchmarking — providing superior model training data than any single syndicate can achieve alone
Yr 2–3
5
Full Agentic Underwriting — Autonomous Bind for Defined Segments
For well-defined, data-rich risk segments below defined TIV thresholds, the system moves toward autonomous bind authority — releasing senior UW for complex, relationship-intensive, and large-line business
Yr 3+
The Agentic AI Architecture
🔗
Data Fabric Layer
CDR, PPL, external APIs (HSE, JBA, Companies House, NCSC), internal loss data, satellite feeds — unified into a real-time risk intelligence data lake
↓
🔵
Guideline Agent
Rule-based + LLM hybrid. 127 Lloyd's/SBP rules enforced in real-time against submission data. Zero exceptions without CUO escalation.
🟡
Synthesis Agent
RAG-enabled LLM cross-references broker narrative against external intelligence. Surfaces contradictions, hidden hazards, and narrative gaps as structured findings.
🟢
Anti-Selection Engine
XGBoost + LightGBM ensemble. 47-feature broker and risk profile. Real-time CDR-integrated loss ratio benchmarking per broker channel and class.
↓
⬡
Composite Decision Engine + NBA Generator
Weighted composite score from all three filters. Priority-ranked Next Best Action prompts. Decision: Approve / Refer / Decline with full rationale. All outputs auditable and explainable.
↓
👤
UW Decision + Override
Human authority preserved. AI outputs are mandatory input. Overrides tracked, attributed, and reviewed quarterly by CUO governance board.
🔄
Feedback Learning Loop
Bind decisions, policy outcomes, and closed loss data auto-feed model retraining pipeline. System intelligence improves with every cycle.
Chapter 09 · Strategic Outcomes
Outcome Horizon: Delivering Sovereignty at Every Stage
Short, medium, and long-term value realisation — from operational discipline through to full underwriting intelligence leadership in the London Market
📊 Click any outcome card to explore in depth
All outcomes are indicative. Realisation depends on deployment maturity, market conditions, and organisational execution.
Short Term · 0–12 Months
Operational Discipline & Visibility
Phase 1 → Phase 2 Completion
100%
Guideline compliance on AI-processed submissions from Month 4
Systematic guideline enforcement eliminates ad-hoc softening with immediate effect from Phase 1 go-live
Broker quality scores established and shared with Active Underwriter — first quantitative basis for placement prioritisation decisions
External data synthesis (HSE, Companies House, HMCTS) live for three pilot classes — 62% of large hidden hazards surfaced at submission stage
UW processing capacity increases 2× on pilot class volume as triage time falls from days to hours
Full audit trail established for SMCR compliance — CUO personal accountability documentation available from Day 1
PMD Decile 10 watch list risk eliminated through consistent enforcement of Lloyd's Minimum Standards
Medium Term · 12–36 Months
Portfolio Quality & Competitive Position
Full Triple Filter Live · Continuous Learning Active
4–6 pts
Combined ratio improvement vs. pre-deployment baseline (Year 2)
Anti-selection detection demonstrably shifts broker channel mix — top-tier broker proportion rises from ~44% toward 60%+ of bound premium
Loss ratio improvement of 8–12% visible in attritional claims on pilot classes, with early signals on large loss frequency reduction
Quote cycle reduced to sub-6 hours on standard submissions — competitive with AI-native carrier benchmarks
Broker behaviour adaptation: lower-tier brokers improve submission quality in response to systematic scoring — a market-level positive externality
Emerging risk signals (climate transition, supply chain, cyber) feeding Filter II in near real-time — ahead of catastrophe model update cycles
Solvency II capital model updated to reflect improved risk selection quality — 4–6% SCR reduction unlocks capital for growth
Long Term · 36 Months+
Underwriting Sovereignty & Market Leadership
Agentic Loop · Cross-Syndicate Benchmarking
Top Quartile
Lloyd's combined ratio ranking — sustained across a full soft cycle
Autonomous triage operational for sub-threshold risks — 70%+ of submission volume handled end-to-end by AI, releasing senior UW for complex and large-line business
Triple Filter becomes a material competitive differentiator in broker conversations — top-tier brokers actively prefer routing quality business to Triple Filter-enabled syndicates
Institutional knowledge fully encoded in the system — talent retirement risk neutralised, underwriting standards perpetuated independent of individual tenure
Lloyd's PMD recognises Triple Filter as model for market-wide AI underwriting governance — potential for Lloyd's-level IP collaboration or licensing
Proven cycle-through performance: combined ratio maintained within 3 points of hard market levels even as the market softens — the most powerful evidence of sovereign underwriting discipline
Cross-class extension complete — Property D&F, Casualty XL, Cyber, Marine Cargo, and Political Risk all covered by class-specific Filter configurations
⬡
The Mandate for the Board
The Triple Filter is not an IT project. It is a strategic decision about the kind of underwriting enterprise you intend to be in 2028 and beyond — one that uses AI to enforce the discipline that the market cycle will attempt to erode, that sees what brokers choose not to show, and that systematically attracts the best risks from the best brokers. The window to act before your principal competitors is now.
Explore
→
1 / 9
Cover
Detail
—