Founder letter
The RooftopOS Reckoning
The market argument behind the company — where the money is running, why the middleman model is failing in public, and what changed when enforcement stopped being a threat. Written from the desk, not from a deck.
Ken Criscione · Founder
I lived every problem this fixes.
I spent two decades at the desk and in the box. I didn’t read about this business in a deck — I lived it, deal by deal, up by up.
And I watched the same thing happen over and over: the middleman moved in. The auction taught the industry to bid against itself. The instant-buyers rolled into our own service drives to poach our own customers’ cars. A stack of vendors wedged themselves between me and my customer and taxed the relationship a little more every renewal. They stopped fighting for the customer and started renting you back to them.
Nobody in that pile ever closed a deal in their life. So I built the thing I always wanted, on my own roof — and I didn’t stop at one. I built three.
He runs an F&I desk in Hartford, Connecticut and continues to write deals while building the product.
That store is RooftopOS’s founding design partner. It is named, and the relationship set out, on the customers page.
RooftopOS publishes no photograph of a dealership: it does not own one, and the store it works inside has not authorised its premises, its signage or its inventory to be pictured.
This was never about software. It’s about people.
Software doesn’t sell a car. A person does. Your customers are the lifeblood — not leads to route out the back door. Your people are the heartbeat — the closer, the tech, the appraiser who deserve a paycheck that grows when they get great. Every incumbent forgot both. We build the other way.
Four ways the middlemen are eating your lunch right now.
The auction mentality is crumbling in public
The purest version of “bid against the dealer” is failing in the open. CarGurus’ board voted on Aug 6, 2025 to wind down all of CarOffer by the end of 2025 — its own words, “less effective in today’s more volatile and unpredictable pricing environment.” Vroom shut its used-car operation in Jan 2024.3,4
Opinion? Sure. But watch where the money’s running — don’t chase it into the ditch.
The instant-buyers hunt your trades — in your own drive
Carvana and CarMax didn’t wait for your customer to come to them. CarMax bought 223,000 vehicles from consumers in FY2025; Carvana moved 596,641 retail units in 2025. They’re not on your side — they’re standing between you and a customer who was already yours.5,6
The dead web form is where good deals go to die
The trade widget on your site hands over a value range and drops a lead in your CRM — then somebody follows up whenever they get to it. An HBR audit of 2,241 companies clocked average first response at 42 hours, 23% never answered. A range is not an offer. A lead is not a close. A form is not a system.1
March 13, 2026 — enforcement stopped being a threat
The FTC warned 97 dealership groups that the advertised price has to be the total price, under Section 5. Put on notice, not charged — but the message is unmistakable, backed by a $20M Leader Automotive judgment and a $3.1M Lindsay payment-packing penalty. It’s not the change. It’s the enforcement.2,8,9
Not a stack of tools. One machine.
Acquire the car, engage the customer, sell it right — every step on your own roof, in your own brand, with your own people. Here’s the family.
Buy your own customer’s car off your own site — a real offer in under 60 seconds.
Read the storyMeet the customer on video — the same conversation that buys the car.
Read the storyA truthful window sticker that sells accessories — every “yes” provable and FTC-aligned.
Read the storyauto(CURB) acquires the car. auto(FILM) engages the customer. auto(LABELS) sells it right. One login, one shared record, one bill — as you connect them.
So I built three. Here’s each one — and why it wins.
Engage them → Earn their trust → Convert them
auto(CURB)
The problem I saw · For 25 years I bought at auction and off-lease — and by the time we reconditioned the car, I owned it above clean trade, closer to retail, watching depreciation eat the margin before it hit the line. At the auction, only the house and the seller win; the buyer always loses. And every “fix” out there was a data company renting me a widget — a dead form with a 42-hour turnaround, or my own lead sold to the dealer down the street.
So I built it · So I built auto(CURB) — buy the car direct from the customer: off-street, off your own website, from for-sale-by-owners, and off your own service drive. 20 seconds to a cash range, 60 seconds to a firm offer. No yes? The AI asks for 5 photos and re-appraises. Still no? Your acquisition manager gets a text and email to close it — and soon, an AI agent that follows up till the deal is done.
Why it wins · The auction inflates what you pay, which inflates what your customer pays. Buy direct and you put better cars on your lot at a lower cost with more margin. And a service-drive car — you serviced it, so you already know what it needs; you’re not blindly buying an auction car that turns out to need brakes, tires, and bodywork.3,4
- Direct from the customer — off-street, your site, FSBO, service drive
- 20-second cash range → 60-second firm offer
- No yes? AI 5-photo re-appraisal
- Then your manager — soon an AI agent — closes it
auto(FILM)
The problem I saw · auto(FILM) grew right out of auto(CURB) — I wanted something more holistic. Your customer lives on video and text now, but this business still hides behind a silent VDP and a phone nobody answers. Plenty of video vendors do a slice of it, but none were built by somebody who has actually closed a deal.
So I built it · auto(FILM) puts your salespeople on camera and lets the customer act right from the video — book an appointment, schedule a test drive, or raise their hand with a trade. Then it carries into service: your customers already expect the MPI video, but auto(FILM) timestamps it, so they jump straight to the spot where the tech shows exactly what the car needs. And it remembers — next visit, it asks if they want to move forward on what they passed on last time.
Why it wins · One experience running sales AND service on the same rail. Every video vendor stops at “communicate.” auto(FILM) turns the conversation into an appointment, a trade, or an approved repair — and the same interaction is built to feed the auto(CURB) appraisal. Video that doesn’t move metal is just content.
- Sell on camera — customer books, test-drives, or trades from the video
- Timestamped MPI — jump to the tech’s exact recommendation
- Remembers declined work and re-asks next visit
- Sales AND service on one rail — built to feed the AutoCurb appraisal
auto(LABELS)
The problem I saw · When COVID cooled off, everyone got competitive the same way — advertise a lower price, then add equipment onto the car. The FTC has always said the advertised price is the total price, and in March they enforced it: a letter to 97 dealer groups, and the industry shuddered. And here’s the part nobody says out loud — when the question comes, the burden of showing what the customer actually saw falls on you. Policies don’t answer it. A binder doesn’t answer it.
So I built it · I did it the honest way for 25 years, so I built the way to prove it. auto(LABELS) puts a truthful addendum sticker on every car listing exactly what’s added and what’s optional, and verifies the price against your own website so the window matches the ad. The customer signs off on what’s already in the price, on every option they choose — tint, PPF, step rails, protection — and on everything they decline — itemized, timestamped, and tied to their signature.
Why it wins · While we’re at it, the disclosure becomes a vehicle passport — comps, market context, the car’s record — because a buyer researches for weeks and a price alone isn’t a story. And that passport lives on into service, for the life of the car. Everyone else sells a static sticker or a generic e-sign; I tie the live price, the products, the disclosures, and the binding consent into one document that matches the window — FTC-aligned by design.2
- Truthful addendum — every add-on and option listed
- Live price verified against your own website
- Signed: what’s included, chosen, and declined
- Becomes a vehicle passport — for the life of the car
Run one. Run two. Run all three.
Every application earns its keep alone and compounds when you stack them. Evaluate one rooftop over sixty measured days — and see exactly what each one costs before you deploy.
One customer. One system. The flywheel no fragmented vendor can spin.
↺ and a deal sold right earns you the next trade
Every stitch of it on your site, in your brand, with your people. A stack of vendors each taxing a slice can’t spin that — they can only fragment it and charge you more next year. That’s not a stack of tools taped together. That’s a machine.
The straight answers, operator to operator.
Do I have to buy all three to get value?+
No. auto(CURB) and auto(LABELS) are live and each stands on its own — run one, run both. auto(FILM) is in early-access pilot now. They’re designed to compound when you stack them, but every platform earns its keep alone.
Is this just another vendor bolted onto my CRM?+
It’s the opposite. I’m the operator these vendors sold to for 25 years. RooftopOS keeps the car, the customer, and the gross on your own roof and in your own brand — instead of routing your customer out the back door to someone who out-bids you.
How does auto(LABELS) make my deals defensible?+
It ties the live price, the add-ons, the disclosures, and the customer’s binding consent into one document that matches the window sticker — so the advertised price is the total price and every “yes” is captured and provable. FTC-aligned by design, not compliance theater.
What makes auto(CURB) different from a trade-in widget?+
A widget hands over a value range and drops a lead for someone to call two days later. auto(CURB) makes a real offer in under 60 seconds with three closing points — AI re-appraisal from photos, then a text to your appraiser to reshape the deal. A range is not an offer; a lead is not a close.
When is auto(FILM) available?+
It’s in early-access pilot — running in production in my own store today, on real repair orders. It’s an end-to-end video experience across sales and service, and the same conversation that engages the customer is built to feed the auto(CURB) appraisal — that cross-product handoff is the integration we’re building now, not something that runs automatically yet. Request a platform demo and we’ll talk about an early-access slot.
Who’s actually behind this?+
Ken Criscione — 25 years at the desk and in the box, still running an F&I office. Not a data company, not a fund. Every one of these platforms exists because I lived the problem it fixes.
The mirror’s big now. Get in your lane first.
The auction mentality is retreating in public. Enforcement is live. Honesty is finally the edge — and the operator who built RooftopOS already ran this play at the desk for twenty years. Stop paying rent on your own customers.
Every hard number above is sourced. Placeholders name the source; final citations link the primary document.
- 142-hr avg first response, 23% never answered — HBR, “The Short Life of Online Sales Leads.”
- 297 dealer groups on notice; advertised = total price — FTC warning letters, Mar 13 2026 (Section 5).
- 3CarOffer wind-down — CarGurus board decision, Aug 6 2025.
- 4Vroom used-car exit — Jan 2024.
- 5CarMax 223,000 consumer purchases — CarMax FY2025 results.
- 6Carvana 596,641 retail units, 2025 — Carvana shareholder letter.
- 8$20M, ~80% unauthorized add-on — FTC v. Leader Automotive Group.
- 9$3.1M payment-packing penalty — FTC / Lindsay Automotive consent order.