Automotive
Dealer networks, warranty and parts data, and plants full of long-lived systems. We build AI that works inside that estate rather than beside it.
Automotive work is rarely blocked by the model. It is blocked by the estate. A dealer management system predates the questions now being asked of it, the warranty platform owns the claim, the parts catalogue has its own rules, and plant systems will outlive several software cycles. We start from where those systems actually are, not where a clean architecture would put them, and we build the integration layer before anything sits on top of it.
The estate is the design constraint
Most of what an automotive AI system touches was bought at different times and will not be replaced to suit it. Some of it exposes a modern API. Some of it exposes a nightly file, and that file is the contract. We treat those differences as design inputs rather than obstacles, and we are explicit about which data can move in real time, which arrives on a schedule, and which should never leave the system that owns it. That decision usually shapes the result more than the choice of model does.
Dealer and aftersales operations
A franchise network is not one organization. Each dealer runs its own business, and the record of a single customer sits split across the dealer system, the manufacturer's own platforms, the service history and the parts side. Useful AI in sales and aftersales has to read across that split and still respect who is entitled to see what. We build the retrieval and the permission model together, so a service advisor gets an answer drawn from everything relevant and nothing the dealer or the manufacturer is not permitted to share.
Warranty and parts data
Warranty is a documentation problem before it is an analytics problem. A claim has to carry evidence that holds up when someone reviews it, and the earliest written account of a fault usually sits in a technician's narrative rather than in a structured field. We build systems that read those narratives, connect them to parts and vehicle records, and keep the reasoning visible so a decision can be defended afterwards. The founders built and run a clinical operations platform where billing and documentation face exactly that standard, and the discipline transfers directly.
Manufacturing and supply chain visibility
Plant and supplier data is usually being generated already and simply never reaches a system that could act on it. The obstacle is rarely the analysis. It is a mixed estate of equipment and vendors with long service lives, sitting on networks that were never designed to hand data outward. The founders run live network and security operations today, so segmentation, exposure and network constraints are design inputs for us rather than a late discovery. We build the pipelines first, then the monitoring and decision support that depend on them.
What we build here
- An integration layer across dealer, warranty, parts and vehicle systems, built to the interfaces those systems actually expose rather than the ones we would prefer.
- Retrieval and copilots for sales, service and aftersales teams, with a permission model that respects the boundary between manufacturer and dealer.
- Warranty and quality systems that read technician narratives, link them to parts and vehicle records, and leave an evidence trail behind every decision.
- Data pipelines that carry plant, supplier and logistics signals into the systems that are supposed to act on them.
- An evaluation suite that tells you when an answer has degraded, and observability that shows where a system spent its time and why.
- Documentation, runbooks and working sessions so your own engineers can operate and extend everything we hand over.
What this is not
This is not a staffed automotive practice, and we point to no manufacturer references, because we have none. Caltor is a small partner-led firm, and what we bring to this sector is engineering discipline proven on systems the founders built and still run: regulated documentation, strict separation between independent operators, and live network and security operations. We do not build vehicle software, embedded control systems or anything that runs on the vehicle itself. Our work sits in the enterprise systems around the vehicle.
Talk to a partner
If the real constraint is a dealer system, a warranty platform or a plant network that everything else has to work around, describe it and we will tell you what we would build against it.
