24 months · 5 phases · 16 activities
AI doesn't hit all seventeen activities at once. The order is causal. FNOL ingests the data claims investigation runs on. Claims data trains underwriting models. Carriers who sequence right compress combined ratio by 3-5 points.
Run the phases in sequence and margin compounds. Run them in parallel and you scale bad inputs.
Personal lines P&C writes $529B annually at 96.7% combined ratio. Seventeen activities span the value chain. Margin lives in underwriting expense and claims handling. Both are pattern-matching at peak.
Claims investigation, customer service, policy issuance, underwriting, FNOL, billing and collections all match patterns. These seven activities represent 65-70% of displaceable spend. AI targets pattern-matching first.
The incumbents have never been more exposed. Margin is concentrated where AI compresses fastest. The 24-month window sorts early movers from late.
Defining condition
Margin concentration in pattern-matching activities creates structural exposure for carriers who don't operate the AI sequence.
$529B
Annual P&C personal lines DWP
17
Value chain activities traced
8
Activities facing AI displacement
65%
Claims investigation displaceable share
Profit pool snapshot · Today
Bar height shows AI-displaceable fraction remaining at this phase. Bars shrink as activities compress.
Each phase produces inputs the next phase consumes. FNOL data feeds claims triage. Claims investigation accuracy trains underwriting models. Underwriting precision drives distribution economics. Run in order and each phase compounds the next.
Run activities in parallel and you scale bad signals. Claims investigation without FNOL data optimizes the wrong severity. Underwriting without claims history prices the wrong risk. Distribution without underwriting accuracy acquires unprofitable book mix.
FNOL moves to conversational intake using voice and photo. Snapsheet, Hi Marley, and Lemonade show the pattern. Four-minute ingestion replaces fifteen-minute phone trees. Data quality improves.
Customer service AI handles policy and billing questions at 60-70% containment. ML triage routes claims by severity and complexity. Routing accuracy lifts 20 points over rules-based assignment.
Margin impact is small in phase one. Capacity freed is the asset. Staff move from intake and routing to claims investigation and adjudication. Phase two consumes this capacity.
What to measure
Four metrics by month six: FNOL time under four minutes, service containment above 60%, triage accuracy lift above 15 points, capacity redeployment rate.
<4 min
FNOL conversational intake target
60-70%
Customer service AI containment
+20pt
Triage routing accuracy lift
$0.5B
Phase 1 cost compression projected
Profit pool snapshot · Months 0-6
Bar height shows AI-displaceable fraction remaining at this phase. Bars shrink as activities compress.
FNOL ingests the photos, voice notes, and damage descriptions claims investigation consumes. Get FNOL right and you capture high-resolution data. Get it wrong and claims investigation optimizes low-quality inputs for six months.
Triage severity depends on FNOL data quality. Photo-based estimation requires photos FNOL captured. Subrogation identification needs accident descriptions FNOL recorded. Start with investigation and you process garbage. Start with FNOL and investigation works on clean signals.
Claims investigation moves to photo-based damage estimation. CCC and Tractable handle auto claims. Cape Analytics and Arturo process property. Virtual handling reaches 60-75% for auto, 40-50% for property.
Subrogation identification runs at adjudication, not months later. Each 1% recovery rate improvement compresses combined ratio by 0.5 points. Recovery rate lifts 10-15% over manual identification.
Fraud detection scores the portfolio at submission and FNOL. Detection lift runs 3-5x over SIU baseline. Better fraud signal feeds faster adjudication feeds higher subrogation lift. The cascade compounds.
What to measure
Combined ratio compression of 1-2 points by month twelve, driven by virtual handling rate, subrogation lift, and fraud detection accuracy.
60-75%
Auto claims handled virtually
+10-15%
Subrogation recovery lift
3-5x
Fraud detection lift vs SIU baseline
1-2pt
Combined ratio compression by month 12
Profit pool snapshot · Months 6-12
Bar height shows AI-displaceable fraction remaining at this phase. Bars shrink as activities compress.
Run claims activities in isolation and each compresses incrementally. Photo estimation saves adjuster time. Subrogation catches a few more recoveries. Fraud detection flags some bad claims. Total impact: 0.5 points combined ratio.
Run as a cascade and the activities reinforce. Fraud scores inform investigation depth. Investigation accuracy identifies subrogation candidates. Subrogation recovery trains fraud models. The loop tightens. Combined ratio drops 1-2 points.
Underwriting moves from rules engines to context-aware risk scoring. Telematics data, property imagery, behavioral signals, and claims history combine. Risk precision improves. Adverse selection compresses.
Auto-bind rate moves from 65-70% to 85-90% on simple risks. Customer service AI handles 60-70% of contacts without transfer. Policy admin compresses as routine work automates.
Policy issuance time drops from days to hours. Endorsement processing moves to self-service at 80%+ rate. Agent and CSR capacity redeploys to complex cases and retention.
What to measure
Cumulative combined ratio compression of 2-3 points through phase three, measured by auto-bind rate, service containment, and book mix quality.
85-90%
Auto-bind rate target
60-70%
Customer service AI containment
80%+
Endorsement self-service rate
2-3pt
Combined ratio compression cumulative
Profit pool snapshot · Months 12-18
Bar height shows AI-displaceable fraction remaining at this phase. Bars shrink as activities compress.
Underwriting changes pricing. Pricing changes which risks bind. Book mix changes loss ratio two years forward. Get underwriting wrong and you write bad risks at good prices or good risks at bad prices for 24 months.
Context-aware scoring requires claims data from phase two and FNOL data from phase one. Skip the sequence and underwriting trains on stale signals. The model compounds old biases. Combined ratio widens instead of compressing.
Distribution compression reaches scale. Direct-channel acquisition costs run 3-5% of premium. Agency costs run 12-15%. Carriers with fast quoting, accurate pricing, and low service costs win direct distribution.
Carriers who haven't operated activity layers in phases one through three face structural margin disadvantage. Their combined ratios stay flat while distribution costs widen. Margin migrates to early movers.
Regulatory reporting compresses from 10 days to 3. Quoting moves from batch overnight to continuous. Re-pricing moves from annual cycles to quarterly or monthly. Carriers adapt or face adverse selection.
What to measure
Cumulative combined ratio spread of 4-7 points between early movers and late, driven by margin compression and distribution economics.
3-5pt
Combined ratio compression for early movers
+1-2pt
Combined ratio drag for late movers
10 → 3 days
Stat accounting close compression
$86B
Total displaceable pool over the chain
Profit pool snapshot · Months 18-24
Bar height shows AI-displaceable fraction remaining at this phase. Bars shrink as activities compress.
Distribution economics force the structural sort. Carriers who operated activity layers see 3-5 point combined ratio compression. Direct CAC at 3-5% vs agency 12-15% compounds the margin. Winners appear.
Carriers who modernized cores without operating activities see flat combined ratios. Their loss ratios stay stable but distribution costs widen as agents migrate to winning carriers. The spread opens and doesn't close.
The playbook is sequential, not parallel. Each phase produces the inputs the next phase consumes. Run them in order and combined ratio compresses 3-5 points. Run them in parallel and you scale bad inputs.
Moative arrives with the thesis written and operates phases one through four as a managed capability. We work the sequence, measure the gates, and compress combined ratio. Outcome-paid.
Sequence is causal. Time is the operating variable.