Infrastructure for Two-Sided Placement
The infrastructure for two-sided market inference
Nexio is embeddable AI infrastructure that evaluates two-sided placement problems and returns ranked candidates with full reasoning, confidence signals, and audit trails. One API call.
import nexio
client = nexio.Client(
api_key="nx_live_k8s2..."
)
# Evaluate a placement in one call
result = client.evaluate(
profile={
"entity": "acme-manufacturing",
"state": "CA",
"line": "professional_liability",
"revenue": 12_000_000
},
pool="carriers:all"
)Placement Evaluation
Hartford Financial Services
Assessment: Strong fit for CA E&O placement
Try It
See the engine process a real scenario
Watch the three-stage pipeline filter, score, and rank candidates in real time. No sign-up required.
Entity Profile
Press Run Evaluation to start
The Problem
Your placement process is a liability
Every platform that makes two-sided placement decisions hits the same wall. The output works until someone asks you to defend it.
01
One score hides everything
You collapsed a multi-dimensional decision into a single number. Your team can’t see the tradeoffs. Your clients can’t understand the output. And when the result is wrong, nobody knows which input caused it.
02
Regulators ask, you improvise
A compliance officer requests documentation on why Client X was placed with Carrier Y. Your team reverse-engineers the reasoning from memory. The audit trail is a person, not a system.
03
Expertise walks out the door
Your best people carry the real evaluation logic in their heads. When they leave, years of domain knowledge leave with them. You can’t scale what you can’t encode.
What You Get Back
Ranked candidates with defensible reasoning
Every response includes tiered candidates, per-dimension scores, natural-language reasoning, and a permanent audit trail link. Not a score. A decision.
{
"eval_id": "eval_7xKp2mNc",
"pool": 1847,
"filtered": 53,
"candidates": [
{
"tier": "STRONG_FIT",
"name": "Hartford Financial",
"confidence": "HIGH",
"scores": {
"appetite": "L1", "coverage": "L2",
"financial": "L1", "pricing": "L1",
"placement": "L1", "service": "L2"
},
"reasoning": "Strong CA prof. liability appetite,
$10-15M band. Preferred pricing tier.
Cyber sublimit gap flagged."
},
// ... additional tiered candidates
],
"audit": "https://api.nexio.dev/traces/eval_7xKp2mNc"
}Three Orders of Intelligence
Hard rules first, then structured inference
Every evaluation is traceable. Every exclusion is justified. The pipeline separates what’s verifiable from what requires judgment.
ORDER 01
Thousands to dozensRelevance
First-order intelligence: deterministic rules eliminate non-viable candidates instantly. Licensing, jurisdictional requirements, appetite flags, regulatory exclusions — binary checks that don’t require inference. Thousands narrow to dozens in milliseconds.
ORDER 02
6 × 4 categorical scorecardInference
Second-order intelligence: multivariate assessment across six independent dimensions. Each scored categorically (L1–L4), not numerically — because “strong appetite fit” is more honest than 78 out of 100. This is where the real intelligence lives: tradeoffs a rules-based system cannot see.
ORDER 03
Full audit trailReasoning
Third-order intelligence: per-candidate reasoning, confidence signals, tradeoff analysis, and complete decision paths. Why Candidate A over B. What was considered. What was ruled out, and why. No black boxes. Every output defensible under audit.
Developer Experience
One endpoint. Structured output. Ship today
REST API with structured JSON responses. Synchronous or async. Renders in any UI. Your users never leave your product.
# Evaluate a placement
curl -X POST https://api.nexio.dev/v1/evaluate \
-H "Authorization: Bearer $NEXIO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"profile": {
"entity": "acme-mfg",
"state": "CA",
"line": "professional_liability"
},
"pool": "carriers:all"
}';
> 200 OK // Tiered candidates + audit trailThe Engine
Every output can withstand an audit
In regulated industries, the evaluation isn’t the hard part. Defending it is. The engine is designed so every output holds up under scrutiny.
Audit Trails
Every evaluation produces a permanent, immutable trace. Input data, inference steps, scoring rationale, and final output — all linked. Hand it to a compliance officer, a regulator, or a courtroom.
Explainability
No black boxes. Every score has a reason. Every exclusion has a rule. Every tradeoff is surfaced. Your team can inspect, challenge, and override with full context.
Human in the Loop
The engine augments expert judgment — it doesn’t replace it. Every recommendation surfaces full reasoning so the human can interrogate, override, and refine. Configurable guardrails enforce hard constraints that inference can’t override.
Learning Loops
The engine gets smarter from outcomes. Placement success rates feed back into assessment calibration. Your domain expertise compounds over time instead of walking out the door.
Verticals
One engine, any two-sided market
Insurance Placement
Evaluate clients against large carrier pools across six independent dimensions. Score appetite fit, coverage alignment, pricing competitiveness, and placement likelihood — each with categorical reasoning. Reduce submission declines. Standardize placement quality across your team.
Mortgage & Lending
Evaluate borrowers against lenders and assistance programs with transparent reasoning across rate, approval likelihood, servicing quality, and program eligibility. Every recommendation defensible under fair lending review.
Enterprise Advisory
Evaluate clients against service providers, consultants, or vendors using structured multi-dimensional assessment. RFP governance, investment analysis, vendor selection — any problem where both sides have complex requirements.
See the engine make a real decision
Walk through a live evaluation. See the reasoning. Inspect the audit trail. No commitment required.