In the last article we said Eric Ries was right about MVPs, and Steve Jobs was right about not shipping the embarrassing version, and the discipline was knowing which mode you were in.
Then a reader asked the obvious follow-up.
What about Elon Musk?
Because Musk doesn’t ship the embarrassing version in private the way Jobs did. He also doesn’t quite ship the Ries MVP either. What he does is something else entirely — and it’s worth naming, because most enterprise leaders don’t realize there’s a whole taxonomy of MVPs out there, each one doing a fundamentally different job.
Lump them all together under the same three-letter label and you’ll make bad decisions about which one to use, when.
Let me lay out the full landscape.
The six kinds of MVP
1. The Smoke Test (or Landing Page MVP). Put up a single web page describing the product, measure how many people sign up, click “buy,” or hand over an email. What you’re testing is interest. Buffer famously launched this way — a landing page that described the product, a button that didn’t lead anywhere yet, and a measurement of how many people clicked it. If nobody clicks, you’ve learned something important before writing a single line of code.
2. The Pre-order / Sell-Before-You-Build. Take real money — or a refundable deposit — before the product exists. What you’re testing is purchase intent, which is a much stronger signal than interest. This is the Musk move. Cybertruck launched in 2019 with a $100 refundable deposit and racked up reservations by the millions. The Tesla Semi was pre-sold to fleet customers like PepsiCo years before any truck shipped. The new Roadster has been on pre-order since 2017. The product wasn’t there. The market signal was.
3. The Concierge MVP. You manually deliver the service yourself to a tiny group of users, and they know it’s manual. Food on the Table famously started this way — the founder personally researched recipes and built grocery lists for individual families before any software was built. What you’re testing is whether anyone actually wants the outcome, regardless of how it gets delivered.
4. The Wizard of Oz MVP. Same as concierge, except users don’t know it’s manual. The product looks fully automated. Behind the curtain, humans are doing all the work. Zappos’s original model was this — Nick Swinmurn photographed shoes at a local store, posted them online for sale, and when someone bought a pair, he went to the store, bought them, and shipped them. The customer thought they were dealing with a shoe warehouse. They were dealing with a guy and a camera. What you’re testing is the full workflow and product-market fit without paying to build the technology.
5. The Piecemeal MVP (sometimes called Frankenstein). You stitch together existing tools — Stripe, Typeform, Zapier, Airtable, off-the-shelf SaaS — and deliver the actual service end-to-end without writing custom code. What you’re testing is whether the end-to-end value proposition holds together, before you commit to building any of it from scratch.
6. The Single-Feature Product. You build one feature really well rather than the full vision in watered-down form. Dropbox launched as “files sync across your computers” — not file sharing, not collaboration, not enterprise admin, not version control. Just sync. Instagram launched as “square photos with filters” — not stories, not reels, not shopping, not DMs. Just filtered photos. What you’re testing is whether the single core thing is valuable enough to build an audience around.
Six different artifacts. Six different jobs. One label that gets used for all of them.
What Musk is actually doing
Look at #2 again, because this is where the most interesting and least-discussed pattern in modern product strategy lives.
When Musk unveiled the Cybertruck in 2019, the product didn’t exist. There were no factories tooled for it. There were no production prototypes. There was a stage prop, a famously broken window demo, and a $100 refundable deposit button on Tesla’s website.
By Monday morning he had over 200,000 reservations. Within weeks, more than 250,000. Within a few years, reportedly over 1.9 million.
The conventional way to read this is “great marketing.” That undersells what was actually happening.
What Musk was doing was running a Sell-Before-You-Build MVP at enormous scale. The product was the question. The reservation was the answer. He wasn’t gathering opinions about whether people wanted a futuristic electric pickup. He was gathering signals — measurable, monetizable, public — that real people would put real money down to be in line for one.
That’s a very different test from what Eric Ries described, and a very different test from what Steve Jobs did with the iPhone reveal. Jobs revealed a finished product. Ries said ship the embarrassing version to real users. Musk did neither. He revealed a concept and asked people to vote with their wallets.
The same pattern shows up across most of Musk’s product launches:
- The original Tesla Roadster (2006) — pre-sold to early adopters with deposits before production lines were ready.
- The Tesla Semi (2017) — unveiled with pre-orders from PepsiCo and others while the truck was still years from delivery.
- The new Roadster (2017) — taking deposits years before production.
- The Cybertruck (2019) — millions of reservations before any consumer had driven one.
The pattern isn’t a stunt. It’s a methodology. You build a vision, attach a deposit, and let the deposits tell you whether the market is there before you tool a factory.
Why this method is so different from “the MVP” most leaders think they know
Most leaders, when they hear “MVP,” reach for one of two mental models.
The first is the Ries model — ship something rough, get user feedback, iterate. Build to learn.
The second, more common in enterprise settings, is a scope-reduction model — ship the smallest version of the product we’re going to build anyway, just to get it out the door faster. This is often called an MVP but isn’t actually testing anything. It’s a project plan.
The Musk model is neither of those. It’s something more like a market sensor than a product. The point isn’t to gather user feedback (the product doesn’t exist yet, so there’s nothing to give feedback on). The point isn’t to ship a smaller version of what you’re going to build (you haven’t committed to building anything yet). The point is to find out if the market is real before you commit capital.
That’s a different test, run against a different audience, with a different decision at the end of it.
If the deposits roll in, you build it. If they don’t, you don’t. The MVP isn’t the product. The MVP is the purchase intent measurement.
When this kind of MVP fits — and when it doesn’t
This is the part that matters most for enterprise leaders, because the Sell-Before-You-Build MVP looks tempting from a distance but is much harder to deploy than it appears.
It fits when:
- You have brand credibility strong enough that people will believe the product will eventually ship. Tesla had that. Most companies don’t.
- The product is concrete enough to be visualized — a vehicle, a device, a tangible offering with a price tag. Abstract enterprise software doesn’t pre-order well.
- The buyer is willing to put something at stake — money, time, a signed letter of intent, a slot on a roadmap. Sentiment doesn’t count.
- The cost of being wrong is high enough to justify a market test before you commit. Tooling a Cybertruck factory costs a lot more than running a smoke-test landing page.
It doesn’t fit when:
- The “deposit” is too soft to mean anything. A $100 refundable reservation measures curiosity, not commitment. So does an internal survey that says “I’d definitely use this AI tool.” Both are real signals, but neither is real demand.
- The product is something the buyer can’t visualize concretely yet. You can pre-sell a pickup truck. You can’t pre-sell “an AI strategy.”
- The buyer doesn’t trust you to deliver. New entrants and unknown brands struggle to convert pre-orders even when the underlying product would be a hit.
The honest version of this method requires distinguishing between the signal of people saying they want something and the signal of people committing something real to get it.
Translating this to AI-Native transformation
Here’s why this matters in 2026 for anyone running AI initiatives inside a Canadian enterprise.
Right now, every leader I work with is drowning in soft signals.
The CEO came back from a conference and said “we should be using AI for X.” Three department heads have said “I’d love a tool that does Y.” A consultant’s report says “the market for Z is going to be huge.” Internal surveys come back showing “73% of employees are interested in AI-assisted workflows.”
None of this is real demand. It’s the enterprise equivalent of a refundable $100 deposit — directionally interesting, but not strong enough to bet a factory on.
What would real demand look like? It would look like a department head signing a service-level agreement that says “if you ship this AI tool, I will commit 30% of my team’s time to using it for six months.” It would look like a budget line moved from one cost center to a new one. It would look like an executive sponsor who’s willing to publicly attach their name to the outcome.
Those are commitments. The rest is curiosity.
The reason the Musk-style Sell-Before-You-Build MVP is so interesting as a lens isn’t because his specific technique transfers directly to your AI program. It’s because it forces a question most enterprise AI programs don’t ask:
Before I build this, what’s the actual commitment from the people who claim to want it?
If the answer is “nothing — they just said it would be cool,” you’re not running an MVP. You’re funding a hypothesis with no skin in the game on the other side. That’s the most common way enterprise AI programs waste money in 2026.
The discipline of knowing which MVP you’re running
Once you can see the six different kinds of MVP, the discipline of running one becomes clearer.
You’re not asking “should we build an MVP.” You’re asking:
- What am I trying to measure? Interest, purchase intent, workflow viability, outcome desire, full end-to-end value, or single-feature pull?
- Which MVP type measures that? A landing page measures interest. A pre-order measures purchase intent. A Wizard of Oz measures workflow viability. A concierge measures outcome desire. A piecemeal measures end-to-end value. A single-feature product measures pull on the core feature.
- What’s the threshold for a green light? Set it before you start. “If we get fewer than X conversions in Y weeks, we don’t build this.” Most enterprises never set this threshold, which is why their MVPs never produce a decision — they just produce more meetings.
- What’s the cost of being wrong in each direction? Building something nobody wants is expensive. Killing something the market actually wanted is also expensive. The MVP type you choose should match which mistake costs more.
That’s not a product framework. That’s a decision framework.
The closing thought
Eric Ries was right. Steve Jobs was right. Elon Musk is doing something different from both of them, and it’s also right, in its context.
The MVP isn’t one thing. It’s six things. Each one is doing a different job. Pick the one that matches the question you’re actually trying to answer.
Most leaders ship “an MVP” without ever naming the question. That’s not a methodology — that’s a label slapped onto whatever they were going to build anyway.
Name the question. Pick the MVP type that answers it. Set the threshold that decides the next step.
That’s the discipline. That’s what every great product story — Bezos, Jobs, Ries, Musk — has in common when you strip the names off.
They knew what they were measuring before they shipped anything.
