Every valid build. One conversation.
SpecSelect is a conversational layer that sits on top of the configurator you already run. Buyers describe what they want in their own words; SpecSelect translates it into a validated configuration and drives your existing visualizer. No re-platforming. No data leaving your infrastructure.
Runs entirely inside your infrastructure · No changes to your existing stack · Governed by your rules
Dialogue on the left. Your visualizer on the right.
Two scenarios, step by step. In the first, the request fits the platform. In the second it doesn't, and that's the interesting case.
The buyer states what they want. SpecSelect checks it against the active parts matrix, applies the options, updates the price and moves the camera.
The buyer asks for a combination this platform doesn't allow. SpecSelect declines, gives the reason, and offers the nearest valid alternative. The price does not move.
Scripted demonstration · brand-neutral reference vehicle
What is SpecSelect?
SpecSelect is a self-hosted conversational layer for vehicle configurators. It reads the existing vehicle catalog, translates natural language into validated configurations, and drives the existing 2D or 3D visualizer through standard events. SpecSelect runs entirely inside the operator's own infrastructure and requires no change to the existing system.
Configurators rarely fail on technology. They fail because a buyer has to know how a manufacturer names its options before they can even begin looking. SpecSelect removes that requirement without touching the system underneath.
What nested option catalogs cost you.
{{ leak.title }}
{{ leak.body }}
Five commitments everything else is built on.
The hard question about an AI assistant on a manufacturer's site isn't whether it impresses. It's whether data leaves the building, whether the existing system has to be touched, and whether the assistant says things it shouldn't. The first three commitments answer those questions, the last two describe what the layer does on top. Hover or tap a card for the full detail.
SpecSelect reads your vehicle catalog through the interfaces you already run and emits standard events to the visualizer you already use. No configurator replacement, no catalog migration, no new CMS. The layer goes on top, not in place of.
SpecSelect runs inside your own environment and holds no buyer data of its own. Conversation history and configuration data used for recommendations sit on your systems. Processing happens where the data already lives. There is no external processing step outside your infrastructure.
A guardrail layer sits between the language model and the configurator. It enforces your internal policies: which combinations are permitted, how pricing may be discussed, what the assistant may and may not say. Invalid configurations never reach the visualizer because they don't get past the guardrail.
Buyers know what they want a car to look like, not what the manufacturer calls its options. Someone describes a car from a film, a look they saw somewhere, a direction. SpecSelect translates that description into options from your catalog.
Where a reference sits outside the catalog, the assistant can issue a style query outbound. That query is structurally constrained and passes the same guardrail as any configuration change: it carries style attributes, never the customer's own wording, never catalog or pricing data. The feature can be switched off entirely per deployment.
From what a buyer has already described, SpecSelect derives options that fit and that they would never have found in the menu. Every recommendation passes the same check as any other change: what isn't permitted on this platform is never proposed.
For the operator, that means premium equipment becomes visible where it previously sat several levels deep.
Four steps from catalog to validated configuration.
{{ s.title }}
{{ s.body }}
Who SpecSelect is being built for.
FACTORY TRIM · CATALOG SPEC
Manufacturers (OEM)
Multi-tier model and trim lines where a buyer has to understand the catalog structure before they can find anything.
BESPOKE CARBON · PAINT-TO-SAMPLE
Tuners and coachbuilders
Individualisation programmes where almost any combination is technically possible and consultants currently explain exclusions by hand.
Retail and dealer platforms
Multi-brand inventory where intent translates badly into filters.
Technical and commercial questions.
{{ f.a }}
Early access for a limited number of partners.
We're working with a small number of manufacturers, coachbuilders and platform operators to test SpecSelect against real catalogs. As a partner you bring a data model and your exclusion rules; you get the conversational layer built on top of them.