Automotive · AI · 2024–Present

YoYo — AI Agent for Dealerships

Multi-tenant chat and voice AI for car dealerships.

Context

The problem.

Dealerships lose most of their inbound leads to slow response times and after-hours silence. YoYo runs as the always-on first line — qualifying the shopper, matching them to live inventory, and only handing off to a human salesperson when the conversation is genuinely sales-ready.

Role

What I owned.

Built and tuned multi-tenant agent flows for chat and voice channels.
Integrated live inventory matching against per-dealer feeds.
Worked across the OpenAI and Anthropic Claude pipelines for routing and tool use.
Hardened conversation memory and lead-handoff into the dealer CRM.

Approach

How it came together.

Multi-tenant from day one

A single agent codebase serves every dealership in the cohort, with per-tenant prompts, inventory connectors, and brand voice. New dealers onboard via configuration, not a forked deployment.

Two model providers, one contract

OpenAI and Anthropic Claude both sit behind the same internal interface, so each conversation can route to the better model for the task — pricing, capability, and reliability tuned per workload.

Voice and chat, same brain

The same lead-qualification logic powers both surfaces. Voice gets the latency and turn-taking treatment it needs; the underlying intent extraction and CRM handoff stay shared.

Outcome

Where it stands.

Live across multiple dealer brands, including the Only Used Tesla storefront.
24/7 lead engagement replacing previously-missed after-hours inquiries.
Conversations cleanly handed off into dealer CRMs with full transcript context.