On July 8 we met with Neptune Flood's CEO Trevor Burgess and engineering lead Jean-Luc Eckstein for a walkthrough of their agent quoting experience, ahead of our API integration. This is the tour for everyone who wasn't in the room: cleaned-up screens, what makes the workflow good, and where it plugs into Brightway.
Neptune is a private flood carrier with an agent-facing quoting portal (their AI copilot, Atlas+, runs alongside it). There is one way to build a quote and three ways it can end. Everything on this page exists in production today — the screenshots further down are just this same flow, step by step.
Hurricane season selling peaks August–October. July is the runway. The shared goal from the meeting: go live soon with something awesome — phases 2–3 can be enhancements over time.
Trevor quoted a real Saint Petersburg property live, from address to bound-ready policy, in a couple of minutes. Faces and meeting chrome cropped out; these are the shared screens only.
Address, aerial map with the parcel outlined, agent attribution, and the live policy cost pinned to the header. Note the error model: instead of blocking, a red Errors (4) counter sits where Bind will eventually be. Atlas+ (their AI copilot) opens with proactive local context: properties in zip 33704 took $143M in flood damage from Hurricane Helene, 830+ claims.
"Give me 3 quick bullets to help sell this policy" → massive local risk, higher limits than NFIP ($250k cap vs Neptune's $4M), 10-day vs 30-day waiting period. Trevor's framing: this "helps the agent feel smart talking about flood" and drives close rates. Quick actions include compare deductibles, recommend coverages, and draft a personalized sales email.
The foundation question — where customers usually stumble — is answered with an illustrated picker (slab, crawlspace, basement, elevated, each with example drawings). Everything defaults sensibly; the agent only refines.
Required fields show inline errors, and the header counter ticks down as they're resolved — Errors (4) → (2) → Bind. The quote and price exist the whole time; completeness only gates the final action, not the exploration.
Switching construction Wood Frame → Masonry updates the header price in place ($7,484 → $7,451). No "recalculate" button, no page reload — the Google Flights lever model we keep referencing.
"Show me some deductible options side-by-side" → a compact table of every tier with premium and delta ($1k = +$2,159 … $25k = −$2,783), each with a one-click Select. This is the most common live-call negotiation (customer asks "what if the deductible were higher?"), served in seconds.
Atlas doesn't stop at the table — it narrates the trade-off (the property sits 6.5 ft below base flood elevation, so first-dollar protection matters) and lands on a stance: keep $5k or go $10k. Then: "want me to apply a change?"
The agent says yes, and Atlas edits the quote itself — "Quote updated — Deductible to $25,000", header reprices to $4,668. This is the agentic step beyond our current Ray patterns: conversational panel + the ability to safely mutate the working quote.
Email the quote and the customer gets a link where they can adjust coverages, watch the premium update, and bind on their own — with the originating agent keeping full credit and commission. Self-service that doesn't disintermediate the agent.
Clean one-pager: property characteristics, coverage schedule A–M, selected deductible, and total annual cost. Familiar shape for anyone who's read a dec page — easy for advisors to walk a customer through.
If the customer declines flood, the agent sends a Flood Rejection email: the insured signs an acknowledgment that coverage was offered and refused — E&O protection with an automatic audit trail. And the email gives them one more chance to change their mind and buy online (~20% do). Also available as a raw form via API if we'd rather run it through our own comms.
"I have been offered this quote … and have declined to purchase this policy. I understand I am not covered for losses associated with flooding." Signature, name, date — plus the full quote details below so the declined offer is unambiguous and recoverable.
Opportunities called out in the meeting, mapped to our surfaces.
Early quote return ("Rate Call 1") means a flood price can appear next to homeowner quotes at address-entry time in Fusion — bundling for attachment rate, refined automatically as home data fills in.
When a renewal case loads for a customer without flood coverage, surface a live Neptune quote as a retention/cross-sell lever — fits the existing Ray ambient-intelligence pattern.
Return the direct-to-consumer link via API and put it in the customer portal — especially during hurricane season when customers log in scared and motivated.
Mirror the rejection email inside our own comms (form available via API), so every declined flood offer produces an E&O-grade record — and a second-chance sale.
25–30k quotes/day + Atlas-style local risk stats ("rates in this zip…") enable zip-code marketing and quote-not-closed follow-ups with actual prices, not generic pitches.
Atlas+ validates our two Ray modes — and adds a third worth designing: copilot actions that modify the working quote ("want me to apply a change?"). Also feeds Rob's sales-scripting tool with talking points.
So design work lands where engineering already is.
flood.policy.effectiveDate, flood.basementCoverage, flood.poolCoverage, flood.hasPriorFloodLoss. Serhii