Lesson 2 mapped the 22-step FNOL-to-reserve pipeline. This lesson is the artifact-level build of Steps 1-7 - FNOL intake through adjuster assignment - on the Atlanta auto BI scenario. The insured at 7:18 a.m. texts Hi Marley: "I had an accident at Peachtree and 10th, the other driver ran a red light, my airbag deployed, I'm at Emory Hospital." By 7:44 a.m. - 26 minutes from intake - the claim is in Guidewire ClaimCenter as 2026-05-AT-08412, Five Sigma has produced the one-page coverage analysis, the adjuster has verified coverage at Step 4 with the §4 reason chain captured, Snapsheet triage routed to complex auto BI specialist, and adjuster J. Martinez is assigned. The 11:00 a.m. first-touch contact closes Step 7 within the 24-hour NAIC Unfair Claims regulatory window with 20 hours of margin. The platforms are Hi Marley (#244 Deloitte Technology Fast 500) for conversational intake across SMS, voice, and web chat; Five Sigma (Starr Insurance 2025 deployment with 40% faster resolution, 35% claims-handling cost reduction, 60% faster general-queue email, 70% accuracy benchmarks; Sutherland partnership for managed services) and Roots Automation for coverage check; Guidewire ClaimCenter (or Duck Creek Claims or Sapiens ClaimsPro or Snapsheet) for claims management; and a workflow engine routing across the 6 swim lanes the carrier operates. The §4 reason chain at Step 4 coverage verification is the highest bad-faith exposure point in the entire pipeline; the handoff design from Lesson 3 enforces the four-element pattern (AI confidence + reasoning + escalation triggers + file note). The California Cumis screen fires here for California operations. Texas §541 documentation discipline anchors here. This lesson is the artifact-level build of those seven steps, the worked Atlanta scenario end-to-end, and the §4.4 documentation packet generated automatically as a byproduct of disciplined pipeline operation.

Step 1 - FNOL Intake via Hi Marley

Hi Marley's conversational AI handles FNOL intake via SMS (the dominant channel for 2026 personal-lines auto FNOL), voice (via the carrier's claim hotline with Hi Marley fronting the IVR), and web chat (carrier portal or mobile app). The architecture: NLP-driven conversation flow with insurance-specific training, structured-field extraction in real time, integration with the carrier's claims-management system at intake completion. Hi Marley's Deloitte Technology Fast 500 ranking at #244 reflects the platform's scale and growth trajectory; the 2026 carrier deployment base spans Tier-1 personal-auto carriers, regional mutuals, and several specialty MGAs.

The Atlanta intake transcript. Hi Marley to the insured: "Hi, this is your insurer's claims assistant. I'm sorry to hear something happened. Can you tell me what occurred?" The insured: "I had an accident at Peachtree and 10th, the other driver ran a red light, my airbag deployed, I'm at Emory Hospital." Hi Marley extracts: claim type (auto BI/PD - the airbag plus the hospital indicates injury), location (Peachtree and 10th, Atlanta), time (extracted from message metadata plus the insured response), parties (the insured plus the other driver), injury indicator (yes - airbag deployment plus hospitalization), severity indicator (significant - Emory Hospital admission rather than urgent-care visit). Hi Marley asks 4-6 follow-up structured questions: vehicle make/model/VIN (cross-referenced against the policy vehicle schedule), occupants in the insured's vehicle, witness contact information if any, police report status, current safety status of the insured and any occupants, preferred contact method for the adjuster follow-up.

The intake closes at 7:18:32 a.m. Hi Marley to the insured: "Thanks. Your claim is being created now. An adjuster will call you within 24 hours. If you need anything urgent, call 1-800-XXX-XXXX. Take care of yourself." The structured intake JSON: { claim_type: 'auto_bi_pd', loss_location: 'Peachtree and 10th St, Atlanta, GA', loss_datetime: '2026-05-28 07:14:00', parties: [{ role: 'insured', vehicle: 'Make/Model/VIN' }, { role: 'other_party', vehicle: 'unknown_at_intake' }], injury_indicator: true, severity_indicator: 'significant_hospitalization', initial_facts: 'Other driver ran red light per insured', police_report_status: 'pending_per_insured', witness_status: 'one_witness_with_contact', communication_preference: 'sms_via_hi_marley' }.

SLA: under 90 seconds for structured intake. The Atlanta scenario completed in 4 minutes including conversational pacing; structured-intake completion at 90 seconds; the remaining time captured additional context (witness, police report, contact preference) that augments the structured record. Hi Marley's conversational pacing is deliberately gentle for injury-indicator intakes; the platform's training data and behavioral guidelines emphasize empathy and pacing over speed when the human is in distress.

Step 2 - Claim Creation in ClaimCenter

Step 2 writes the structured intake JSON into Guidewire ClaimCenter (or Duck Creek Claims, Sapiens ClaimsPro, Snapsheet - carrier-dependent). The claim number assigns from sequence. The loss-date timestamps. The NAIC Unfair Claims Settlement Practices Act acknowledgment SLA clock starts.

Atlanta Step 2. 7:20 a.m. ClaimCenter creates claim 2026-05-AT-08412. Loss date 2026-05-28 07:14:00. The policy lookup confirms the named insured matches the policy holder; the vehicle VIN matches the policy-listed vehicle (no schedule mismatch that would trigger coverage analysis at Step 3); the coverage period is current; the premium is paid through the next billing cycle; no policy modifications are pending. The claim file populates with the intake data plus the policy data. The UCSPA acknowledgment timestamp is captured. SLA: under 2 minutes from intake completion; the Atlanta claim completed in 2 minutes.

Step 3 - Coverage Check via Five Sigma

Five Sigma reads the policy declarations, endorsements, and exclusions; produces a one-page coverage analysis. Roots Automation's cognitive intake worker provides an alternative platform with comparable capability - most carriers run one or the other rather than both. Five Sigma's Starr Insurance 2025 deployment produced the published benchmarks: 40% faster resolution, 35% claims-handling cost reduction, 60% faster general-queue email response, 70% accuracy on coverage determinations validated against expert review. The Sutherland partnership announced in 2026 provides managed-services support for carriers deploying Five Sigma at scale.

Atlanta coverage analysis. Five Sigma output: covered (the named insured is driving the insured vehicle); BI limits 100/300; PD limit 50; deductible $500 PD; comprehensive deductible $500; collision deductible $500; UM/UIM 100/300 elected; medical payments $5,000 elected; no exclusions triggered (the commute use disclosed in the declarations matches the accident location at the intersection during the typical commute time); rental car coverage $30/day to $900 max; towing and labor coverage $100. Coverage status: full coverage available with no immediate reservation-of-rights candidates identified. SLA: under 4 minutes; Atlanta completed in 4 minutes.

The Five Sigma vs. Roots architectural difference. Five Sigma: ML-driven coverage analysis with carrier-specific policy schema mapping; strong on standard auto and property coverage forms; weaker on commercial complex coverage forms with multiple manuscript endorsements. Roots Automation: a cognitive intake worker that functions like a digital co-worker reading policy documents end to end; better on complex commercial coverage forms with multiple endorsements and modifications; carries higher cost per claim. Mid-2026 split: 55% Five Sigma, 30% Roots Automation, 15% other or in-platform tools (Guidewire's coverage module, Duck Creek's coverage module). For Atlanta personal auto, Five Sigma is the appropriate choice; the standard auto form maps cleanly into Five Sigma's schema and the 70% accuracy benchmark is high enough for the adjuster's verification load to remain manageable.

Step 4 - Coverage Verification by Adjuster

Step 4 is the §4 reason-code checkpoint with the highest bad-faith exposure in the entire pipeline. The adjuster reads Five Sigma's analysis, opens the actual policy in the PAS, verifies the AI's analysis against the policy declarations and the policy text, signs off on the coverage status. Override authority is captured in the reason chain. The Texas §541 documentation discipline anchors here; the California §790.03 fair-claims practice anchors here; the Florida §624.155 CRN posture anchors here; the New York Insurance Law §2601 unfair claims posture anchors here.

Atlanta Step 4. 7:36 a.m. Adjuster J. Martinez reviews the Five Sigma summary. Opens the policy in ClaimCenter. Confirms: the named insured matches the FNOL contact, the vehicle VIN matches the policy schedule, the accident date is within the policy effective period, no business-use exclusion concern (the commute use is disclosed and the accident occurred at an intersection during typical commute time at a typical commute location), no driver exclusion (the insured is a named driver per the policy declarations), no late notice issue (the notice was received same day as loss). The adjuster signs off on coverage at 7:36 a.m. §4 reason chain: "AI Five Sigma analysis verified. Coverage available per BI 100/300, PD 50 with $500 deductible. No exclusions triggered. No reservation-of-rights candidates. Adjuster verification complete. California Cumis screen: not applicable (Georgia loss, no California operations relevance). Texas §541 documentation discipline: complete. Five Sigma invocation ID 2026-05-28-V8412 captured." Adjuster signs.

The four-element handoff per Lesson 3. AI confidence on coverage: 96% (high-confidence due to the standard coverage facts and the clean policy lookup). Reasoning: 5 verification factors confirmed (named insured, vehicle VIN, accident date, business-use, driver exclusion, late notice). Escalation triggers evaluated: coverage gray-area (not triggered - coverage clear), reservation-of-rights candidate (not triggered - no exclusion concern), Cumis screen (not applicable - Georgia loss), business-use exclusion (not triggered - commute disclosed). File note draft completed; the adjuster signs. Total adjuster time: 12 minutes; SLA: under 15 minutes; on target.

The high-stakes failure mode at Step 4. An adjuster who rubber-stamps the Five Sigma analysis without opening the policy is the carrier's worst-case bad-faith vulnerability. If the policy contains a business-use exclusion that Five Sigma missed (the schema mapping was outdated for that policy form's edition; the endorsement library lagged a 2025 carrier-specific modification), and the adjuster did not verify, and the claim is later denied on a business-use basis, the discoverable file note shows the adjuster accepted the AI analysis without independent verification. Texas §541 punitive damages and Florida §624.155 CRN exposure both compound in that fact pattern. The 12-minute Step 4 SLA is calibrated to allow substantive verification; the discipline is that the SLA exists to enable verification, not to compress past it.

Step 5 - Triage Routing - Fast-Track vs. Complex

Snapsheet triage (or Roots Automation, or in-platform triage in ClaimCenter) scores the claim against the carrier's triage rules. Fast-track: clear coverage, no injury, single-party loss under $7,500 (auto) - claims that resolve in the carrier's automated estimating and settlement workflow with minimal adjuster touch. Complex: injury, multi-party, large-loss, coverage gray-area, prior-claim density, SIU referral indicators.

Atlanta Step 5 - complex routing. 7:40 a.m. Snapsheet evaluates: injury indicator present (the claimant at Emory Hospital with airbag deployment); multi-party loss (insured plus other driver, with the other-driver liability still to be confirmed); ambiguous liability ("other driver ran red light" is the insured's assertion, not independently verified pending the police report); coverage clear (Step 4 verified). Triage rule 5.3.2 (injury plus multi-party) routes complex. SLA: under 4 minutes; completed in 4 minutes.

The §4 reason chain at Step 5. Triage decision: complex. Drivers: injury_indicator (Emory hospitalization plus airbag deployment), multi-party_loss (2 vehicles), ambiguous_liability (claimant assertion not yet verified by police report). Alternative path considered: fast-track rejected because the injury indicator triggers complex routing regardless of the dollar magnitude. Reason chain references: triage_rule_5_3_2_injury_multi_party, reserve_trajectory_monitor_5_3_4_emory_indicator (signals potential large-loss BI exposure that requires the senior adjuster's experience).

Step 6 - Adjuster Assignment

The workload-balancing engine assigns to the right desk. Complex auto BI routes to a senior adjuster with auto-BI specialty. The carrier maintains specialty sub-lanes: auto BI, property large-loss, WC, L&H disability, BI demand-package, fraud-focused. The sub-lane assignment determines which adjuster's experience and authority match the claim's complexity profile.

Atlanta Step 6. 7:44 a.m. ClaimCenter's workload engine assigns to J. Martinez (auto BI specialist, current caseload 18 active claims, capacity 22, expertise tag auto_bi_complex, jurisdiction Georgia). The adjuster receives the assignment notification with the claim summary. Estimated complexity score: medium-high based on injury plus multi-party plus ambiguous liability. SLA: under 4 minutes; completed in 4 minutes. The assignment includes the §4 reason chain audit trail from Steps 1-5 so the adjuster opens the file with full context.

Step 7 - First-Touch Contact SLA Enforcement

NAIC UCSPA plus state DOI bulletins require first-touch contact within 24 hours for most lines, 12 hours for catastrophe. The adjuster contacts the insured. AI drafts the first-touch script and the recorded-statement outline; the adjuster makes the call. The first-touch quality determines downstream cooperation, settlement trajectory, and any later claims-handling complaint posture.

Atlanta Step 7 - 11:00 a.m. first-touch call. Adjuster J. Martinez calls the insured. AI-drafted script opens: "Hi, this is J. Martinez with [carrier]. I'm so sorry about the accident. I understand you're at Emory - how are you feeling right now? I want to walk you through the next steps." The conversation captures: current medical status (admitted, awaiting MRI; concussion plus bruised ribs suspected), expected hospital stay (overnight observation), other driver contact (the insured has the phone number from the scene exchange), police report status (officer at scene; report number obtained; release expected 24-48 hours per the Atlanta PD standard cycle), witness information (one witness with contact info), preferred follow-up communication (text message Hi Marley channel), preferences for medical bill submission (direct to carrier vs. claimant submission), legal representation status (none currently - no attorney letter received). 18-minute call. SLA: 24-hour first-touch window; achieved within 4 hours of FNOL.

The recorded statement. AI offers the recorded-statement option: "Would it be helpful to take a brief recorded statement now to capture the basic facts while they're fresh? It's voluntary." The insured agrees. Hi Marley records the statement (with the insured's explicit consent captured on the recording); speech-to-text transcribes; Roots Automation summarizes into structured fields aligned to the carrier's recorded-statement template. The recorded statement transcript and the structured summary attach to the claim file by end of business day. SLA on transcription: under 24 hours from the call.

The §4 reason chain at Step 7. First-touch achieved within 4 hours of FNOL; UCSPA 24-hour SLA met with 20 hours of margin. AI-drafted script reviewed and customized by adjuster before call. 18-minute conversation captured with structured field extraction. Recorded statement obtained with explicit consent. Legal-representation status confirmed (none); the carrier's communication may continue directly with the insured rather than through counsel. Next-step plan documented in file note signed by J. Martinez.

The 26-Minute Intake-to-Assignment Summary

Steps 1-6 elapsed time for the Atlanta scenario: 26 minutes from FNOL to adjuster assignment with coverage verified and triage routing complete. Pre-AI baseline at typical carriers: 2-6 hours for the same sequence depending on staffing density and the FNOL channel. AI-enabled compression: 80-90% on intake-to-assignment.

Step 7 first-touch at 11:00 a.m. closes the regulatory window with 20 hours of margin. The recorded statement transcription completes by 4:00 p.m. closing Step 8. The pipeline continues into investigation, photo damage assessment (Tractable at Step 10 - covered in Lesson 8), reserve recommendation (Step 13), medical-records review when records arrive (Step 17 - covered in Lesson 10 under the HIPAA §164.502 minimum-necessary discipline), and the parallel fraud screen at Step 15 (Lesson 9).

The cycle-time delta as competitive moat. The 80-90% compression on intake-to-assignment is itself a competitive advantage. Insureds who experience the smooth conversational FNOL, the 4-hour first-touch, and the structured-statement experience perceive the carrier as professional and responsive; the customer-satisfaction effect on retention is measurable (carriers running Hi Marley plus Five Sigma plus Snapsheet stack report 6-12 point NPS improvement on claims experience versus comparable carriers running legacy phone-IVR intake). The retention impact compounds across the book.

The §4.4 Documentation Packet at End of Step 7

By 4:00 p.m. Day 1 of the Atlanta claim, the §4.4 documentation packet contains:

Hi Marley intake transcript with timestamps and the conversational pacing record. The ClaimCenter claim record with the policy lookup confirmation. The Five Sigma coverage analysis JSON output with the invocation ID and model version. The adjuster Step 4 verification reason chain with the four-element handoff structure (confidence 96%, 5-factor reasoning, escalation triggers evaluated, file note signed). The Snapsheet triage decision with the reason chain (complex routing, drivers list, alternative path rejection). The adjuster assignment record (J. Martinez, caseload, expertise tags, jurisdiction). The first-touch call recording plus the AI summary. The recorded statement transcript plus the structured summary. The UCSPA acknowledgment timestamp. The first-touch SLA achievement timestamp with the 20-hour margin documented.

Ten artifacts in the first-day §4.4 packet. A §4 examiner six months later can reconstruct the intake-to-assignment sequence with full attribution: which AI ran, what it produced, what the adjuster verified, what the supervisor approved, what the regulator's UCSPA clock measured. The handoff design enforces the documentation; the platforms produce it automatically; the carrier captures the audit trail for free as a byproduct of disciplined pipeline operation. The 2026 best-practice carrier extends the same discipline to the L&H, commercial property, and commercial GL lines so the §4.4 packet template is consistent across the claims organization.

Key Takeaways

• Hi Marley conversational intake handles SMS + voice + web chat with NLP-driven flow. Atlanta scenario: insured texts at 7:18 a.m.; the structured intake JSON completes at 90 seconds; the conversational pacing extends to 4 minutes including witness, police report, contact preference capture. Hi Marley's Deloitte Technology Fast 500 ranking at #244 reflects platform scale.
• Step 2 ClaimCenter claim creation under 2 minutes from intake completion. Claim 2026-05-AT-08412; loss date 7:14 a.m.; the NAIC UCSPA acknowledgment SLA clock starts. The policy lookup confirms the named insured, vehicle VIN, coverage period, and premium status.
• Step 3 Five Sigma coverage analysis under 4 minutes. Output: covered, BI 100/300, PD 50 with $500 deductible, comprehensive + collision $500, UM/UIM elected, medical payments $5K, no exclusions triggered, no reservation-of-rights candidates. Five Sigma's Starr Insurance benchmarks: 40% faster resolution, 35% cost reduction, 60% faster email response, 70% accuracy. Sutherland partnership for managed services in 2026.
• Step 4 adjuster verification under 15 minutes - highest bad-faith exposure point. J. Martinez reviews Five Sigma, opens the policy, confirms 5 verification factors (named insured, vehicle VIN, accident date, no business-use exclusion, no driver exclusion, no late notice). §4 reason chain captures the 4-element handoff: 96% AI confidence, reasoning, escalation triggers evaluated, file note signed. Texas §541, California §790.03, Florida §624.155, New York §2601 documentation disciplines all anchor here.
• Step 5 Snapsheet triage routes complex. Triggers: injury indicator (Emory plus airbag), multi-party loss, ambiguous liability (insured assertion not yet verified). Triage rule 5.3.2 and reserve_trajectory_monitor_5_3_4 fire. Alternative fast-track path rejected because the injury triggers complex routing regardless of dollar magnitude.
• Step 6 ClaimCenter assigns J. Martinez (auto BI specialist, 18/22 caseload, expertise tag auto_bi_complex, jurisdiction Georgia). Estimated complexity medium-high based on injury plus multi-party plus ambiguous liability. The §4 reason chain audit trail from Steps 1-5 ships with the assignment so the adjuster opens the file with full context.
• Step 7 first-touch call at 11:00 a.m. - 20 hours within the 24-hour regulatory window. AI-drafted script reviewed by adjuster. 18-minute call captures medical status, other driver contact, police report status, witness, communication preferences, legal representation status. Recorded statement obtained with explicit consent and transcribed by 4:00 p.m. (Step 8 closes).
• Steps 1-6 elapsed time: 26 minutes intake to assignment. Pre-AI baseline 2-6 hours. Compression 80-90%. §4.4 documentation packet contains 10 artifacts by end of business day reconstructing full intake-to-assignment sequence. The cycle-time delta is itself a retention advantage - carriers running the Hi Marley + Five Sigma + Snapsheet stack report 6-12 point NPS improvement on claims experience.