Pull your last software invoice stack. Not the DMS — everything around it. CRM, chat, equity mining, marketing automation, the AI add-on your CRM vendor started charging for last year. Now multiply by sixty months. That's the bill for renting a brain you'll never own.
Public ballparks — and treat these as ballparks, because every store's stack is different — put a franchise dealership's total technology spend north of $15,000 per rooftop per month once the DMS is counted, spread across seven to ten vendors. Strip out the DMS and the transactional plumbing you genuinely can't replace, and the slice that's up for grabs — CRM seats, AI chat, lead handling, reactivation tools, marketing automation, the point solutions that each do one thing — commonly runs $6,000–$8,000 per rooftop per month. Even small businesses far outside auto retail average $110,000+ on AI SaaS over five years and hold nothing at the end (we covered that number in insights no. 01). A dealer group runs that math with more zeros.
Rent buys usage. It never buys the asset.
The five-year math, on one page
Take a three-rooftop group and hold the numbers conservative. Everything below is an illustration — the audit runs it on your actual invoices.
- The rent path. $7,000 per rooftop per month on the replaceable slice × 3 rooftops = $21,000/mo. Over five years that's $1.26M — before the 5–8% annual price increases most SaaS contracts carry, which push the real figure toward $1.4M. At month sixty you own: nothing. Cancel any line and that capability — and the data inside it — walks out the door.
- The own path. A build phase measured in months, not years, at our published engagement range ($10–30k/mo while we stand the system up), then a managed per-rooftop license priced to come in under the stack it replaces — that's the point of replacing it. Assume, illustratively, the license lands at half the rent it retires: the five-year total comes out around $650–750k. At month sixty you own: the system, the configuration, the data, and the off-switch.
Same five years. Roughly half the cash. And the halves aren't equivalent, because one path ends with an asset on your balance sheet and the other ends with a renewal notice. The economics moved this direction fast: the price of a fixed amount of AI capability fell roughly 280× in eighteen months, and open models on hardware you control run inference in cents. The premium you pay a per-seat vendor is no longer for the intelligence. It's for the packaging.
The part the invoice doesn't show
Cash is the smaller half of the argument. The bigger half is what compounds.
Every lead your team answers, every deal your desk works, every service visit and trade-in and unsold customer — that history is the raw material of an AI system that actually knows your store. On the rent path, that context sits fragmented across vendor databases, training their product, priced back to you per seat. On the own path it accumulates in one system, on your infrastructure, getting more valuable every month your team uses it. A competitor can buy the same software you rent tomorrow morning. They cannot buy fifteen years of your customer history compounding inside a system you own.
That's the asymmetry that decides this. When you rent, switching costs work against you — the vendor holds your data, so leaving is expensive, so the price rises at renewal, so you stay. When you own, switching costs work for you — the system trained on your context is the thing a competitor can't replicate, and nobody can reprice it against you. Same force, opposite beneficiary. Pick the side of it you want to be on.
The honest risks — both directions
Owning has real risks, and pretending otherwise would be a sales pitch, not a diagnosis. Who maintains it? What happens when models improve? Who answers at 2am? That's exactly why the model is a managed owned instance: we run it, patch it, and improve it — and you hold the keys, the data, and the right to walk with all of it. Management is a service you can re-bid. Ownership isn't.
Now weigh the renting risks, which rarely make the slide deck: your vendor gets acquired and sunsets the product. The per-seat price rises 8% at every renewal because leaving is painful by design. The AI features you fund with your subscription — trained partly on your usage — ship to your competitor the same day they ship to you. And if you ever leave, your data comes back as a CSV export, if it comes back at all.
None of this requires the software to be bad. Most of it is quite good. It's the structure that's bad — for you. Per-seat rental is a magnificent business model, which is precisely why you're on the paying end of it at seven to ten companies simultaneously.
When renting still wins
Fair's fair: if you're a single point running lean, if your team won't use what you have now, or if a process is broken, ownership fixes nothing — an owned system on a broken process just industrializes the dysfunction. The math above starts closing hard for multi-rooftop operators with real lead volume and a customer file worth compounding. If that's not you yet, keep renting and revisit in a year. If it is you, every renewal you sign is a year of asset-building you gave to someone else's balance sheet.
Run it on your numbers
The illustration above is deliberately conservative and deliberately generic. Your stack, your rooftop count, and your lead volume produce your version of it — and that's a one-week exercise, not a leap of faith. Bring the invoice stack to the deployment audit and we'll build the five-year model on your actual spend: what's replaceable, what isn't, where the crossover lands, and what the owned system recovers along the way. You keep the model either way.