Newgate Systems Start a conversation

Case study · AI product

Fourteen hours of searching,
done in fourteen minutes.

Car shopping is fragmented by design. Car Finder replaces it with a conversation that learns what you want, checks what you can afford, and comes back with cars that fit both.

Client
Car Finder
Industry
Automotive retail technology
Region
United States
Platform
Conversational AI, deep search, financing integration

01

The client

The average American buyer spends over fourteen hours searching across multiple car websites, and by one AutoTrader study 89% of buyers feel overwhelmed by how fragmented the process is.

The problem is not a lack of inventory data. It is that every site dumps inventory on the buyer and leaves them to work out which of it is relevant, affordable and financeable.

02

The problem

Existing car sites optimise for listing volume. A buyer arriving with a real-world constraint — a monthly payment ceiling, a family of five, a need for something they can park in a city — has to translate that themselves into filters, then repeat it site by site.

Financing makes it worse. Sticker price is not the number that matters; the monthly payment is. A search that ignores financing shows people cars they cannot buy.

03

Requirements

  • R1Conversational interface that collects preferences progressively, in natural language
  • R2Deep search across multiple inventory databases simultaneously
  • R3Preference learning that refines recommendations over time
  • R4Budget captured early and enforced on every recommendation
  • R5Real financing estimates — payment, deposit, rate, total cost
  • R6Side-by-side comparison and continued background searching

04

The solution

We built the product around the conversation rather than the catalogue. Preferences are collected the way a person would ask for them, and financing is treated as a first-class filter rather than a detail revealed at the end.

The design principle: never show a car the buyer cannot afford. Budget is captured early and applied to everything after it, which is why the results are short — and why they are useful.

05

The work

  • 01Conversational AI. A natural-language assistant that collects preferences progressively and adapts to the user's history.
  • 02Deep search. Simultaneous search across multiple inventory databases.
  • 03Preference learning. Recommendations refined as the system learns the buyer's priorities.
  • 04Financial integration. Payment, deposit, interest rate and total-cost breakdowns attached to every option.
  • 05Comparison tools. Side-by-side vehicle comparison with specifications and pricing.
  • 06Background search. Continued searching on the buyer's behalf as inventory changes.
Conversational AIDeep searchFinancing integrationRecommendation systems

06

Outcome

14 min

Target time to a matched car, from 14 hours of searching

Budget-first

Every recommendation inside the buyer's actual affordability

Learning

Recommendations improve as preferences and inventory change

More work

Other records.

Next

Facing something like this?