Twenty-four marques.
Every dealer principal
wants it yesterday.
A marketing operating system for the whole group. It reads live inventory and turns it into campaigns, holds every marque's CI, runs the creative queue, watches Meta and GA4 and moves before you ask, records the DP meetings and comes back with ideas, and plans the events. It drafts everything. A person approves everything.
Four of the eight run on this page right now. Paste your own stock list, your own ad copy, your own meeting notes. It is the real model, not a recording.
The car is a 3D scan of a 1958 Porsche 356 — the model Colin Giltrap once co-owned one of. Forty thousand points, lit and rotating. A nod, nothing more.
Not a chatbot. The desk that runs group retail marketing.
Industry reporting for 2026 puts the dividing line between dealers who use AI sporadically and those who embed it in the core workflow. This is the second thing. Every capability below reads real sources, drafts real work, and stops for a named person.
This is what came out of the stock list this morning.
Every one of these was triggered by something that changed in inventory, not by a marketing calendar. Note what each one deliberately does not say — and that two of them are campaigns the system refused to run.
Creative is drawn rather than photographed — a concept should not ship imagery it does not own, and generic body-type silhouettes avoid using any marque's trade dress. In a pilot these render against your own approved photography from the asset library, in the right CI for each marque.
The stock list is the best campaign brief you own.
Every campaign idea below came from something that changed in inventory this morning. Nothing was invented, and no idea outlives the car it was about — when the vehicle sells, the campaign retires itself.
Dealer investment in Google Performance Max rose roughly 47% year-on-year to April 2026, with reported conversions up 119% and cost per lead down over a third — industry reporting, not a Giltrap figure. Those gains go to whoever feeds the platforms the freshest inventory signal. That is a data-plumbing problem before it is a media problem.
Paste real stock. Get real campaigns.
Sample rows are loaded below — edit them, or paste your own. The marque sets the register. This runs the actual model and drafts against the rules; nothing is pre-written.
Nothing is stored. No pricing is generated. Every draft comes back needing a named approver.
Give it copy you would actually send.
The sample below has four real problems in it, of the kind that survive a copy-paste at 4pm. Replace it with your own draft if you would rather test it properly.
Indicative only, and not legal advice. A real deployment loads your own claim register and distributor packs.
Every DP meeting ends with a promise. This is what keeps it.
Twenty-plus principals, each rightly focused on their own marque, each wanting it done now. The failure is never goodwill — it is that a Tuesday conversation becomes a note nobody actions, and by the next meeting you are both starting again.
Press play to see what the meeting turns into.
Paste any meeting notes. Get the work out.
Scrappy notes are fine — that is the point. It separates what was decided from what someone has to do, gives every action an owner, flags what is blocked, and adds proactive ideas nobody asked for.
Nothing is stored and nothing is sent. In a deployment this is consent-gated and the follow-up still waits for a person.
Ask it for an activation. It will give you three and a run sheet.
Pick a marque and a moment, add anything you know, and it proposes activations grounded in the marque rather than generic event ideas — plus the bit that always slips, which is the follow-up.
It will not book, invite, or commit spend. Venues and suppliers are suggestions for a person to choose.
No migration. No new system for anybody to learn.
The reason group marketing tools fail is not capability, it is installation. This one is deliberately designed to be switched on beside what you already run, and to be switched off again without leaving a hole.
A scheduled export
Whatever DMS you run already produces stock, sales and service reports. Point one of them at a folder on a schedule. That is the integration.
A folder of brand assets
Group guidelines and each distributor's CI pack, as they already exist. The system reads them as rules and versions them, so the current pack is always the one in use.
Two read-only logins
Meta and GA4, read-only. It watches and recommends; it never spends. Media stays exactly where it is today, with the people who run it today.
Everything else — the creative queue, the meeting notes, the events calendar, the campaign ideas — needs nothing from IT at all. If it is not earning its place in six weeks, the export stops and nothing else changes.
A general model will write you a Fair Trading breach and not mention it.
A system drafting across twenty-four marques, on live inventory, into paid media, is exactly where a compliance problem enters unnoticed. Every draft is checked, and each check names the rule it holds.
Under the Fair Trading Act 1986 a company faces penalties of up to $600,000 per breach, and an individual up to $200,000. (Motor Trade Association, dealer legal guidance.)
Press run to check today's drafts against the rules that actually bite in New Zealand motor retail.
The screen the group does not currently have.
Illustrative sites and work — a fictional Thursday, not Giltrap data. No production access is requested by this concept.
| Site · marque | Work | Source | Waiting on | Status |
|---|
Every row is a draft waiting for a named person. The system does not approve, publish, book media, commit spend, or move money.
What this will never do
It does not send. It drafts, and a named person sends. It does not publish to any channel, write back to your DMS, submit to a distributor portal, book or buy media, commit co-op funds, move money, or price a vehicle. It never touches customer finance data. It records a meeting only with everyone's consent, and any participant can stop it. It does not pretend to be a person — every draft says it was prepared by a machine and approved by a human.
No production access is requested by this concept.
Scope
Two marques and one dealer principal, for six weeks. Live-inventory campaign ideas, the CI guard, and the meeting loop. Not media buying, not the whole group.
Access
One scheduled stock export, your group brand guidelines and two distributor CI packs, and read-only Meta and GA4. No DMS write access. No customer data. No spend authority.
Scorecard
Drafts accepted without rewrite. Campaign ideas a DP actually ran. Compliance flags a person agreed with. Time from a DP asking to the work landing. And the one that matters: would that dealer principal be annoyed if you switched it off?
Fail any line of the scorecard and we change the design or stop.
Pick a verb.
Not “book a demo”. Any of these is a real next step, and the third one is a perfectly good answer.
What is the one constraint we have got wrong?
Every concept is built from the outside. There is always something about how Giltrap actually runs that we could not see. One line is enough.
Opens your mail app to assembl@assembl.co.nz. Nothing is collected by this page.