Fleet vehicle keying, from days to minutes.
claveteoAI takes the broker's file in whatever format it arrives, normalizes its columns, and assigns every vehicle a code from your carrier's catalog, with a score and reasoning in plain language. What it resolves with certainty is auto-validated; the doubtful arrives flagged for review, never in silence.
Today, one person codes the broker's file unit by unit.
Brokers send heterogeneous, incomplete files. Without automation, the Issuance team cannot scale.
claveteoAI in one sentence
A web platform that automates fleet vehicle keying: it receives the broker's Excel file in any format, normalizes the data to the insurer's schema, assigns the correct vehicle key, auto-validates what is safe and sends what is uncertain to the reviewer, keeping an immutable record of every decision.
The catalog is the heart of the system.
Without a normalized catalog, the engine cannot operate.
The system's base input
- Unique ID per vehicle: the insurer's key
- Make, sub-brand, version, year, universe
- Textual model description
- Versioning: every change creates a new snapshot
Connecting to understand
- Vehicles connected with their variants and aliases
- The graph detects brand → model → version relations
- Rescues keys even with partial data
- Base for semantic search via pgvector and embeddings
From Excel file to validated key.
A clear, auditable pipeline, not a magic button. The insurer's catalog feeds the match at every step.
Upload and normalize
The broker uploads the Excel in any format; an LLM maps the headers to the insurer's schema.
Pre-classify and match
Regex, LLM, and VIN pre-classify each record; SQL + pgvector + LLM compare it against the insurer's catalog.
Human review
Only records below the configurable threshold go to a specialist for review.
Audited and immutable
Every decision —automatic or human— is validated and logged in an auditable history.
What changes for the Issuance team.
An honest comparison between today's manual process and the AI-assisted pipeline.
One person against a hundred thousand keys
- ✕ 4-5 days to close a mid-size fleet
- ✕ Manual row-by-row search against the Excel catalog
- ✕ No traceability of why a key was chosen
- ✕ Every correction reprocesses the entire file
- ✕ Errors hard to detect until they reach the rate
One person supervises. The AI closes.
- ✓ Under 5 minutes end-to-end per fleet
- ✓ Auto-validation of everything above the threshold
- ✓ LLM reasoning auditable per record
- ✓ Reprocessing only of modified records
- ✓ Immutable history for regulatory audit
The numbers that matter.
Results measured in the local POC with 200 vehicles. Not yet a large-scale production average.
The same team, multiplied by the pipeline.
Your team stops searching for keys row by row and starts supervising a self-validating pipeline, with a triple anti-hallucination gate and an immutable audit trail on every decision.
Run the discovery with rockyFrom the broker's spreadsheet to a quotable fleet.
Thirty seconds with the real system: upload the file exactly as it arrives, review key, score, and reasoning per vehicle, and resolve the doubtful ones with a single keystroke.
Let's get claveteoAI running with your catalog.
Tell us your fleet size and your vehicle catalog. We'll show you the match working with your data.