Frayze wasn't hired by a dealer group. It was built inside one.

0. The origin: one rep, one satellite store, year one

Frayze didn't start as an agency. It started as a tool I built for myself.

I was working in sales and finance out of one of Need A Car's smaller satellite stores. Even when I was tied up with a deal on the floor, I wanted leads moving — calls returned, customers nurtured, appointments booked. Nothing on the market did what I needed at the speed I needed it. So I started building. Year one was about the local team, the local store, and proving the system could outperform the manual approach without anyone particularly noticing.

1. Online Auto Sales — blanketing 6 stores and 6 languages

After year one, the local store and team were running. The next move was to expand the operation across all six Need A Car rooftops as a unified online sales arm — so a customer landing on the website at 11pm could be worked by the same engine that worked the showroom floor at noon. The wrinkle: this isn't a small-town dealer group. Need A Car serves the GTA, which means real customer demand in six languages: English, Spanish, Arabic, Farsi, French, Hindi, and Tagalog.

• Multi-lingual lead routing: every inbound lead was language-tagged and routed to a salesperson who could close in that language.
• Per-store + group-level pipelines: a customer who reached out via the group site could be claimed and worked by the nearest rooftop, with full handoff history.
• Single-operator-scale tooling: the same system that ran one rep was now running an online sales arm — without changing the underlying logic.

2. The rollout that wasn't supposed to happen — 72 staff in 10 days

I started recruiting one salesperson from each store into the portal as a quiet pilot. Within weeks, the people running on the system were sitting #1 on the board in every city they worked in. Average reps were turning into top performers, on the same lead flow, against the same competition. That got noticed — not always in a flattering way. People got upset. Leadership had a choice: kill the pilot, or roll the whole group onto it.

They chose the rollout. I had 10 days to onboard 72 staff across all 6 locations. No beta. No staging. Live, with hundreds of thousands of contact records flowing in on day one. The only way that was possible was because AI and automation were doing the heavy lifting from the first hour:

• Speed-to-lead automation already worked at scale — every inbound lead got responded to in seconds, not minutes.
• Missed-call text-back on every store line caught what humans couldn't during the chaos.
• Routing logic by language, store, and vehicle interest kept the right lead in front of the right rep without manual triage.
• Live cutover across all rooftops simultaneously — no store ran on two systems for any period.

The rollout held. Within weeks, the bottleneck moved from "can we onboard everyone" to "can we feed everyone enough leads to work." Which is what triggered Phase 3.

3. The AI layer — finding 1 car out of 700, before AI was cool

With 72 reps working the system across a rolling 700-vehicle inventory, the next bottleneck wasn't lead capture — it was matching the right car to the right approved buyer fast enough. We started layering in AI for the parts of the deal that were repetitive but high-stakes:

• Document collection: AI agents requested, classified, and chased customer documents (license, pay stubs, insurance proof) so reps weren't doing it manually for every deal.
• Approval-based vehicle matching: scrapers + AI cross-referenced bank approval terms with the live 700-car inventory to surface the cars a given customer could actually drive home today.
• AI-written vehicle descriptions: better, more search-friendly listings written at inventory scale, kept fresh as stock turned over.
• AI reply layer: on inbound conversations, the system could answer common pre-qualification questions instantly — handing off to a rep at exactly the moment the lead was ready to talk numbers.
• AI voice agents: evaluated and integrated voice-AI stacks (ElevenLabs for voice synthesis, Vapi for conversational voice for a stretch) so the system could place outbound calls, qualify, and book — not just send SMS and email.

4. The owned stack — new site, SQL inventory, feed distribution

The next thing we hit was infrastructure. The existing website had flaws, and Pixel Guru — the third-party photo vendor — was breaking across multiple stores: listings going stale, photos missing, prices out of sync with the floor. So we stopped renting:

• Brand new Need A Car site, built in-house on WordPress with deep customization — designed for conversion, not just listing. VDPs with strong CTAs, finance pre-qualification, lead capture wherever it earned its place.
• SQL-level inventory management — single source of truth for every car, every photo, every price, every status across all six stores. Replaced the legacy listing-management workflow entirely.
• Photo vendor migration: Pixel Guru → CarMedia, with the SQL layer enforcing the standard so a missing photo at any store surfaced immediately instead of being discovered when the listing went stale.
• Feed distribution to every channel that mattered: AutoTrader, CarGurus, Kijiji, Facebook (Marketplace + Ads), Google Merchant Center — fed automatically from the SQL layer, with conversion data flowing back for attribution.
• Per-rep landing pages and business cards with their own SEO and blogs — every salesperson became a small inbound channel.
• End-to-end tracking: ad click → web session → form fill → CRM record → showroom appointment → sold, all stitched.

Phase 4 is where the +400% inbound lead growth and 600–700% monthly traffic increaseshowed up — and where the dealer group stopped being dependent on third-party listing platforms to make a phone ring.

5. First-of-its-kind: AutoCorp 2-way + retention agents

With the owned stack solid, we went deeper into the systems most agencies will never touch:

• 2-way AutoCorp integration (first of its kind): assigning a deal in the CRM reflected instantly in the AutoCorp portal, and vice versa — pipeline stages, owner, status — synced live across both systems. No double-entry, no end-of-day reconciliation. Frayze is now an official AutoCorp partner on the back of this work — the integration is supported, productized, and available to other dealer groups.
• AutoCorp BookIt: consolidated identity, insurance, and license verification onto AutoCorp BookIt — replacing the earlier AutoVerify (Optimize) layer once AutoCorp shipped the equivalent inside their suite. Fewer vendors, tighter integration, lower run-rate cost.
• Sold-vehicle retention agents: automated follow-up for delivered customers — equity check-ins, service reminders, lease-end re-engagement — designed to keep the customer with the group through their next purchase.
• In-house financing payment-collection agents: automated touchpoints for in-house finance customers to keep payments on track, reduce delinquency, and route harder cases to a human at exactly the right escalation moment.
• Built but never rolled: an HR portal and a vehicle-intake / inventory-management portal — scoped, designed, sitting on the shelf for whenever leadership wants to flip them on.

Phase 5 is where the $1.7M annual marketing savings came from. Once cost-per-sold-vehicle could be measured by source — across the SQL inventory layer, AutoCorp, and the CRM — the group could stop spending into vendors and campaigns that couldn't prove a delivered car, and concentrate on the ones that could.

6. The intelligence layer — TV dashboards, showroom screens, ownership view

The system was generating more data than any single human could review. So we built the visibility layer for ownership, management, and the sales floor itself:

• Showroom appointment screens: a customer walking into a store could see their name and the salesperson they were meeting with — small detail, large credibility lift.
• Management heads-up TVs: rotating displays showing live calls, texts, emails, social posts, appointments, missing photos, average days in stock — at a glance, across every rooftop.
• Close-rate and expected-sales dashboards: based on every active lead across every store, in real time, rolled up for ownership.
• Resolution + reputation management: social comment responders, review requests and responders — all coordinated rather than left to individual reps.
• Sponsored-post integrations: direct social-media publishing for paid pushes, tied back to attribution.

7. The platform underneath — every CRM, every chatbot, every vendor

This is the part most case studies skip: the cost-of-ownership trajectory of the actual stack. Over four years we ran, evaluated, and eventually replaced every layer at least once. Each migration was driven by the previous layer costing more than it was returning. The honest history:

• Dealer CRM (legacy): SM360 — the existing dealer CRM the group inherited. Reliable for what it did, but a closed system that couldn't carry the automation, AI, or feed work we needed underneath it.
• CRM evaluation + adoption sequence: VinSolutions → SalesPype → Matador → GoHighLevel + custom. The group migrated off SM360 in stages, with the final stack settling on GoHighLevel as the base with a custom application layer on top for everything GHL alone couldn't do. We kept evaluating alternatives after the system worked — none of them beat the GHL + custom combination.
• Chatbot layer: Tidio (off-the-shelf) → in-house. Off-the-shelf got us moving; in-house was the only thing that could speak to the CRM, AI layer, and language router at the same time.
• Photo / imaging vendor: Pixel Guru → CarMedia, with the SQL layer enforcing the standard (covered in Phase 4).
• Orchestration: Zapier (v1, fast to ship) → direct API calls + webhooks (cost + reliability) → n8n + custom apps (full control over workflows that no SaaS layer could express). Each migration cut a meaningful chunk out of the per-seat / per-task tax.
• Comms layer: Twilio for SMS and voice, Mailgun and SendGrid for transactional email — picked per channel based on cost-per-message and deliverability, not vendor convenience.
• Cloud infrastructure: AWS, with Lambda functions running the scrapers, webhooks, and integration glue. Custom code where the lift was small; managed services where the cost made sense.
• Scrapers when APIs weren't open: we're a dealer-side operator, not a tier-1 vendor — which means several industry APIs are gatekept behind partner agreements we don't qualify for. So we wrote scrapers. Carfax is the clean example: a direct API was prohibitively expensive for a single dealer group's use case, so we built a scraper that delivered the same data at a fraction of the cost. Same approach for several other closed-data sources where the math didn't work the other way.

Phase 7 is where the 52% tech-cost reduction compounded. Fewer tools paying overlapping rents, more workflow owned in-house, and a clean discipline: off-the-shelf when it was cheaper than building, built when the off-the-shelf math stopped working, scraped when the gate was locked and the cost of legitimacy outweighed the value.

8. The moonshot — sell a car start-to-finish, entirely by AI

Underneath all eight phases was one north-star goal: a customer lands on the site, qualifies, picks a car, gets approved, signs, and schedules delivery — without a human touching the deal until they shake hands at pickup. Every component to make that real exists somewhere in the dealer-tech ecosystem. They just don't talk to each other unless you make them.

What we built toward it:

• CreditApp lending-portal integration — pre-qualification and credit-app submission designed to run through the system, not a third-party iframe. We built working staging models end-to-end; production integration was scoped but never shipped before the engagement transitioned.
• Dealertrack scraper — dual purpose: the obvious purpose was pulling deal status and approvals out of Dealertrack and back into the CRM. The less obvious purpose: sales staff often won't use the CRM consistently, no matter how well it's built. The scraper guaranteed that every deal worked through Dealertrack got logged and updated, regardless of whether the rep remembered to put it in the CRM. It was insurance against the most common dealer-group data-integrity failure.
• AutoVerify (Optimize) → AutoCorp BookIt: identity, insurance, and license verification ran on AutoVerify Optimize until AutoCorp shipped equivalent functionality in BookIt. We consolidated onto BookIt — one fewer vendor, tighter integration with the AutoCorp 2-way layer.
• AutoCorp 2-way — already covered in Phase 5. Frayze is now an official AutoCorp partner on the strength of this build.
• DeskIt — desking integration for finalized payment, menu, and product structures.

The one that blocked it:

• Direct Dealertrack API access — never obtained. Dealertrack's API gatekeeping is well-known in the dealer industry; they protect the wall around their tooling, and a dealer-side operator isn't on the partner list. The scraper got us roughly 80% of what direct API access would have delivered. The final 20% — clean handoff of approved deals straight into the F&I desking flow — was where the moonshot stalled.

Behind the system: what 4+ years of planning actually looks like

Every phase above was mapped before it was built. Below are three views of the strategy work — exported from the live planning boards used across the engagement. Most agency case studies don't show the planning because there wasn't any. We're showing it because it's the part that compounds.

These aren't post-hoc diagrams drawn for a case study — they're the working artifacts the build was scoped against. Available on request during a fit call for any dealer group serious about a phased engagement.

Why this matters for your dealer group

Most agencies that sell to dealer groups have never sat in an F&I office, never closed a deal in two languages, never owned a 10-day live rollout. We have. The Need A Car system wasn't a project Frayze pitched to a client — it was the system Frayze built from inside, deal by deal, year by year, until it had earned the right to be productized.

If any of the following sound familiar in your group, the same sequence applies:

• Your CRM is really a different CRM at each store, plus spreadsheets, plus the AutoCorp portal that no one's actually keeping in sync.
• You're paying lead vendors and listing platforms but can't prove cost-per-sold-vehicle by source.
• Your inventory feed, your website, and your photo vendor are owned by three different vendors who don't talk.
• Your database is your biggest asset and your most ignored channel.
• Your reps are getting beat by reps at smaller groups who happen to have better tooling.
• You serve a multi-lingual customer base and your current system pretends you don't.

Want a fit call?

Twenty minutes. We'll map your current stack, identify which of the seven phases you're actually in, and tell you honestly whether a phased engagement makes sense for your group. If you want, we'll walk you through one of the planning boards above on the call.