Tesla designed the Model 3 line to be the most automated in the history of car manufacturing, laid the whole thing out digitally, and installed it before a brick of the factory was built. The line never went into production.
The company survived by doing the opposite of what it had planned: a tent outside the factory, cars assembled by hand, 100 a week and then 500.
"We said automate last. you've got to perfect the process before you automate otherwise you just might bury yourself and in that case we almost did."
Jon McNeill spent three years as president of Tesla reporting directly to Elon Musk, was later chief operating officer of Lyft, and has written the book that sets out the five-step method — question every requirement, delete every step, simplify, accelerate, automate last — in the order Tesla and SpaceX apply it.
The full interview is covered here so you can skip it. 36 minutes of audio, 22 minutes of reading.
Here are the 15 lessons that matter.
👤 Guest: Jon McNeill, Co-Founder and CEO of DVx Ventures, previously President of Tesla and COO of Lyft, and author of The Algorithm, the first book by one of Elon Musk's direct reports
🎙️ Host: Nicolai Tangen, CEO of Norges Bank Investment Management, which runs Norway's sovereign wealth fund
📰 Published: 16 September 2026 on YouTube (In Good Company with Nicolai Tangen)
🔴 YouTube | ⏱️ 36 min | ✅ Time saved: 14 min
Key Takeaways
Tesla's most automated line ever built never produced a car, and the rescue was a tent full of people assembling by hand
The rule that came out of the postmortem is that automation goes last, after the process is already good
A Tesla lawyer found that not one paragraph of a 12-page auto loan document was required by law or regulation
Tesla cut its online configurator from 300,000 possible builds to two models, and the head of manufacturing said it made his life 10x easier
Toyota turned aluminum into a finished car in 4 days against Tesla's 15, which is 2.5x less working capital
Putting AI on an unfixed process makes you wrong faster and more expensively, because you are paying for tokens
Mapping a process on sticky notes and circling only what the customer pays for suggests about 90% of the steps are unnecessary
Tesla aimed at Apple's 30% gross margin against a car industry at 10%, and ran at 24–28% for years
He hired for three things — curiosity, a bias to action and intelligence — and tested the first two with a hard problem in the interview
His biggest career mistakes all came from not using his own company's product
1. Question Every Requirement
Tangen took McNeill through the five steps in order, starting with the one McNeill said unlocks all the others.
The target is anything nobody has interrogated for a long time. People who have looked at the same problem for years stop testing the assumptions underneath it, and many of them, he said, "fall by the wayside as unproven."
The worked example is Tesla's online configurator in 2016. "So like we were trying to sell 100,000 euro cars online for the first time anybody had done this in 2016 and every person that's in e-commerce knows that the more clicks you have the less conversion you have to the actual sale. We had 64 clicks when we started out."
The assumption he questioned was build-to-order itself. Quantified, the configurator produced "over 300,000 different combinations" that a first-stage manufacturer was being asked to build, which he said makes life very hard.
His replacement came from the sales data: "And what the data says is people really buy two cars from us. They buy a performance car or they buy a long range car." Colors stayed configurable; the drivetrain combinations did not.
He had been at the company about a month, and expected to be attacked for it. The head of manufacturing asked what it would do to him. "You would make my life 10x easier."
On why rule-breaking is hard, his answer is about people, not organizations: "I think number one, humans are natural complicators, not simplifiers."
He reached for Mark Twain: "Mark Twain had the famous line, I would have written you a shorter letter if I would have taken the time." And then the point underneath it — "It takes work to simplify and very smart people, very bright people tend to complicate rather than simplify."
2. The 12-Page Car Loan
Tangen asked what was supposed to be untouchable. McNeill named auto financing, which he called one of the worst and one of the most complex parts of the customer journey.
The starting point is a document he said runs "about 12 pages long" around the world, with "dozens and dozens of paragraphs."
He put a single question to the company's lawyer. "So one day I questioned our lawyer and I said why do we need these 12-page documents? How many of these paragraphs are the requirement of law or regulation?"
The lawyer had never been asked. "That's a great question. Nobody's asked that before. Let me come back to you."
The answer was "Precisely none." And then the lawyer went further: "We have all the case law in place to back us up. If somebody doesn't make a payment on a car, we can go get the car."
McNeill asked whether the agreement could be a single paragraph naming the price, the rate, the term and the monthly payment. The transcript records the lawyer answering that they could not, and the conversation moved on before the reason was given.
His test for separating a rule worth keeping from a dumb one is three words long. Rules that make sense are "a requirement of safety, of law, or physics." "Everything else is in the dumb category until proven otherwise."
3. Delete Every Step
Step two is a physical exercise rather than an analytical one.
Managers were made to map the entire process on a wall in sticky notes — one note per step, sub-steps underneath — whether the process was manufacturing, sales or loan processing.
Then one question: which of these steps does the customer pay for. Customers do not pay for quality checks, he said: "They don't pay us for purchase orders. They pay us for the product."
The result is the number that makes the exercise worth doing: "When you do that, it looks like about 90% of the steps may not be necessary."
The specific deletion he described is the quality check between stages, where work would sit waiting for an inspector and then sit waiting again to rejoin the flow. The responsibility moved to the person doing the work.
The control that replaced inspection is measurement of rejects: "And we said to the person downstream, if it's not of high quality, you can pass it back and we're measuring the passbacks and so we can tell where the good quality is and where it isn't."
Tangen raised the Musk line that the best part is no part, and pushed on it — with no parts there is no business, so where did they cut too far and have to add back? McNeill said there were plenty of those, and that having to add back was itself the signal they had gone too far.
His reading of the phrase is combination rather than subtraction. "What that means is can we combine things?" An electric car has a cooling system for the battery and another for the cabin. "Teslas today have one single heat pump system that cools the battery and cools the cabin rather than two. That's one less system that can break and one less system that we have to manufacture."
4. Simplify, Then Optimize
Tangen asked how simplifying differs from deleting. McNeill's answer is that step three is where the shortened process gets run for the first time — by hand.
The instruction is to resist software. "This is hard for technologists because we all want to put hands on keyboards especially in the age of AI. We want to rush to the digital solution. And what we insist teams do is they manually run the process first."
He rejects the usual trade-off outright. The usual formulation is that you can have good, fast or cheap and should pick two. "Pick two. It turns out that really great process yields good fast and cheap." His mechanism: a process has to be high quality to run fast, and running fast at quality means high throughput, which is what makes it cheap.
Asked when simple is simple enough, he gave a financial answer rather than a process one. The goal was margins at twice the industry's.
The benchmark was a technology company. He said the goal was gross margins equal to Apple's: "And Apple's gross margins are roughly 30%. The car industry's gross margins are roughly 10."
What they got: "And so we aimed for 30 and said if we can reach this, we're going to be world class. and for years, the gross margin hovered between 24 and 28%." His comparison: "So we were more than almost two and a half times the gross margin of our competitors."
On why a financial target and not a process one: "We had a hard time defining what perfect looked like, but we knew what great looked like." The metric had to be one the markets and investors could read.
5. Accelerate Cycle Time
Tangen's question was why speed arrives so late in a method associated with a company known for speed.
The reason is that speed multiplies whatever the process already is. "Basically, because you have to if you speed up a bad process, you're just getting to the bad answer faster."
Speed is also the diagnostic. Faults surface first, because it is hard to run fast with faults in the process; remove them and the speed comes.
The Model 3 and Model Y ramp is the illustration: "So in the example of like the model 3 or model Y when we started production we wanted to get 50 cars a week through the production line. Then we sped it up to 100. So we doubled it. Then we doubled it again to 200. Then more than doubled it to 500. Then doubled it again to a thousand."
Asked how you actually make people move faster, he named two levers. The first is waiting time: "most process speed gets lost in downtime," things sitting between steps, and a speed goal finds those first.
The second is repetition. He said people build muscle memory from practice and get better as they repeat the work, and that "machines actually get better with repetition too."
Tangen offered a Formula 1 line about speed as a unifying force, attributed in the conversation to Michael Schumacher, though neither could place it with confidence. McNeill took the phrase, calling speed "the unifying force in almost any process."
6. Velocity of Cash
Asked why it is so hard to get people to hurry, McNeill said the answer is a metric, and that he learned it from Japan.
"The people at Toyota talk about a very different financial metric than the rest of the people in the industry. They talk about velocity of cash and everybody in that organization is wound around when we take a dollar in, how fast can we turn that into a dollar profit."
The comparison is the hardest number in the interview. Tesla took 15 days to go from a pile of aluminum to a finished Model 3 or Model Y — he corrected himself from five to 15 as he said it. Toyota took four.
"So what that means is Toyota needs two and a half times less working capital than we needed." He said he has carried the same metric into the businesses he runs now, and calls velocity of cash the highest level of competition in business.
Tangen asked how China made the same thing a national mindset. McNeill's answer is that China went to school on Japan next door, and started with brute force.
The brute force is the 996 schedule, which Tangen asked him to define: "9:00 a.m. to 9:00 p.m. 6 days a week. That's brute force." Tangen noted it would not pass labor law in many other countries, which McNeill accepted.
The sequence after that is the same algorithm: perfect the process, then automate, with speed goals at every stage. The result he cited is construction time: Chinese contractors built factories for Tesla in "less than half the time" it took in Europe or North America.
7. The Weekly Heartbeat
The operating rhythm McNeill ran at Tesla came from a mentor's line about how to hit a number.
"If you want to make a quarter, make your month. If you want to make your month, make your week." He said the full version continues down to the day and the hour, and that a weekly cadence was as far as he was willing to take it.
The point of the cadence was answerability. Musk could ask at any moment whether the quarter would be made, and McNeill could answer with certainty, because he knew the pulse of the business.
What the meeting actually did: "And so every week that weekly heartbeat was pulling together the demand side of the business and the supply side and making sure that we were absolutely in sync."
He named the shocks it was built to absorb — tariffs and supply problems — and said the intent was to make the quarter one way or another, and exceed it where possible.
8. Automate Last
The last step of the algorithm is the one Tesla learned by nearly destroying itself.
His image for it is irreversibility: "This is last because automation is like a concrete that you pour over a process and once you do to remove it takes a jackhammer." He said you have to be "very careful when you pour that automation in."
He was explicit that the steps were learned from mistakes and postmortems, and that production hell was largely of Tesla's own making.
"We had designed the most automated manufacturing line in the history of automotive manufacturing. And we designed it entirely digitally and we designed the machines digitally and laid out the factory digitally and put all the automation in place before a single brick was laid in the factory."
The failure became visible on a walk of the factory floor with Musk. "Look at these machines. They have to be calibrated every hour." The machines sat six inches apart, leaving no room for a person with tools. "We've designed this digitally. We didn't design it in the real world."
The line never went into production, and the consequence was financial: "We almost went bankrupt because we didn't have the cash flow that we had predicted coming off a Model 3. And the only way we saved ourselves was to go back to the manual process. We literally built a tent in the factory outside and produced cars by hand. First 100 a week, then 500 a week, etc."
The postmortem produced the rule. "We said automate last. you've got to perfect the process before you automate otherwise you just might bury yourself and in that case we almost did."
His example of doing it right is DoorDash, which he said its Stanford founders launched with "PDFs of menus and a telephone number" at the bottom of the screen rather than software, picking up and paying for orders themselves while they mapped the workflow. Tangen noted the show has had the DoorDash founder on.
"Go manual before you go automation because it's going to teach you everything you need to know about the business." The moment to automate, on his account, is when the process is as good as you can make it and already running fast.
9. AI on a Bad Process
Tangen asked what McNeill makes of companies layering AI on top of old, cumbersome processes.
"My thoughts is this is just speeding up disaster."
The cost is doubled rather than merely wasted: "This is ubiquitous with AI right now. People throw AI into an existing process. And you're not only getting to the bad answer faster, you're getting to the bad answer more expensively because you're spending tokens."
His counter-argument to the delay objection is that fixing the process is not the slow part — he said it does not take long and does not take much work.
What he wants instead is targeting. He described challenging executives to find the key levers in their business and apply AI and automation to those, so that the work produces "a P&L impact that they can point to" — which he said matters to the organization and to whoever is funding it.
10. Innovation and Control
Tangen asked whether a method aimed at faster and cheaper actually produces innovation.
McNeill said that was the point of it. The algorithm was used to drive innovation, he said, and "not incremental change but quantum change."
His prediction about how Musk will eventually be assessed is about cadence, not vision. "One of the things that I think academics when they study Elon Musk 20 or 30 years from now and say what made this person such an effective industrialist, one of the things that's going to stand out is that he managed the key aspects of the simplification and innovation of the business weekly and drove weekly progress which adds up over time to look like huge breakthroughs."
Broken down, those breakthroughs are a percent a week until you work out "how to land a rocket and catch it" — or produce a car at twice a competitor's margin, or get the car to drive itself.
Tangen pushed on whether the framework needs full ownership and no labor unions. McNeill took the ownership half: "Yeah the framework definitely favors those who control their entire production system and you'll see that there's a lot of vertical integration once you start to innovate this way because you need to have control of the systems."
His current example is robotics. "A good example of that is in robotics today." Many of the actuators needed to make a robot hand work, he said, "don't actually exist" — so the only route through is to "produce your own and vertically integrate."
He said the same pattern shows up in fast-innovating Chinese manufacturers, naming BYD alongside Tesla.
11. Unrealistic Goals
Tangen brought up the reality distortion field associated with Steve Jobs and asked how deliberately absurd goals fit the method.
McNeill said goal setting deserves its own chapter and that he would add one given a redo.
The mechanism is that a modest goal is achievable with the existing system and a large one is not. "When you set a goal of 5 to 10% growth, you're going to get 5 to 10% growth. When you set a goal of 50 to 100% growth, people have to rethink entirely how they're doing that."
"And so part of Elon's magic and part of Steve Jobs magic with the reality distortion field is to set incredibly ambitious goals" — not only as financial targets, he said, but to change how the people doing the work think.
Tangen asked where it meets realism, using living on Mars in two years as the test case. McNeill's answer is that the goal is priced for partial delivery: "I used to tease Elon that when he put a goal out there, I knew that if we hit 50 or 60% of the goal, he'd be thrilled." He said Musk agreed.
The constraint on ambition is psychological: "It can't be so ridiculous that people just give up from the start and say this is impossible. it's got to be somewhat within reach."
His argument for why starting matters more than the target is compounding. Teams that begin start to learn, the feedback loop folds back on itself, and the result is "recursive learning within an organization" — which builds "a compounding advantage versus your competitors" who have not started.
12. Special Forces
Tangen asked how much fear there was inside the organization. McNeill said the answer surprises people.
"I would say that the most common trait of people at SpaceX or Tesla is humility, believe it or not. And there's not much fear."
The standard response to an impossible goal is two responses in sequence. First humility — "I have no idea how to do that. I have no idea how to achieve that." Then confidence: "we'll go figure it out." He summarized it as "I don't know how to do this, but let's go figure it out."
The second requirement is tolerance for failed attempts — failed launches and failed tests before the goal is reached — absorbed quickly and not repeated. The compounding argument returns here: competitors, he said, are usually "too fearful to start that journey."
On sleeping at the plant: "I spent probably weeks worth of nights on the factory floor sort of for the launch of Model X, for the launch of Model 3." The reason was visibility — being "in the problem with the people" — and teaching, since he and his colleagues considered themselves teachers of the method.
Tangen asked whether his wife approved. She did not, and teased him: "I'm pretty sure the executives at GM or Toyota are not sleeping on the floor."
On burnout, McNeill conceded the model is not sustainable as a permanent state and described how he framed it in hiring. He told recruits "you are joining special forces, not regular army."
The distinction is deployment pattern. Special forces are not deployed continuously: "They're deployed in short bursts of time for key missions." The compensation is the team: "But the trade-off for that is you're going to be working in a platoon with the best of the best. You're going to do the best work of your life."
The ordinary week was not the intense one: "But most nights, if you came into the office at 7:00 p.m., you could roll a bowling ball through this office and not hit anybody." Training ran 8 to 6, and he agreed with Tangen that "you'd burn people out if that was continuous."
13. Hiring for Curiosity
Asked how to spot someone who can execute rather than talk, McNeill named three ingredients and the interview method for testing two of them.
"the first thing we look for is curiosity." Candidates were given a very difficult problem in the interview and watched for how curiosity shaped the way they broke it down and pursued an answer, then asked for examples of doing the same thing before.
The reason curiosity comes first is a claim about temperament. Curious people tend "not to be satisfied with the status quo."
"Second ingredient we look for is a bias to action." They asked candidates for something from their career or studies they considered world class, then took it apart — how the insight came, and what first steps were taken.
The third is intelligence, and the three together were the filter. A candidate with "curiosity, bias to action, and intelligence" was, he said, a pretty good bet to thrive at a place like SpaceX or Tesla.
14. The Money Machine
Tangen asked what McNeill does first if he is dropped into a struggling company.
"First thing I do is I look for how the financial model or how the financial machine of the business actually works." What he is after is the two, three or four levers that decide the outcome, because those say where innovation can break something open.
His instruction to management is blunt: "teach me the money machine of the business" and say where the leverage sits. Then the work is doubling, tripling or quadrupling one of them.
The signal that a business is beyond repair is not financial. It is people who accept the status quo rather than being dissatisfied, curious and inclined to act — culture, in other words, is the first thing that would make him leave.
"I think changing culture can happen over time but it takes in my experience long periods of time to change a culture."
His explanation for the persistence is founder DNA: "it's embedded in the DNA of the business," and "most culture comes from founders." His example is General Motors, a company of about 150 years, where he said the character its early leader installed is still visible.
Changing culture, he said, is "not for the faint of heart" and not for anyone short on time.
Tangen offered his own version: "Sometimes I ask people to define the corporate culture and they come up with some defining characteristics and then I just say but hey these are just your personality traits and that's exactly what they are." McNeill agreed — "They're personality traits of the founder."
15. Eat Your Own Dog Food
Asked for the biggest mistake of his life, McNeill limited it to business and named one habit.
The mistake is not using the product: "I would say eating my own dog food. There have been businesses where I have gotten lazy and not sampled the product every day."
The correction came from a book. "And it wasn't until I read Sam Walton's book, Made in America, that I started to appreciate sampling your own product on a daily basis." "And one of the things that Sam says in that book is he famously would call his customer service telephone number every day on his way into work to understand how customers were being greeted and treated."
"And the biggest mistakes I've made is when I've stepped away from the product and I haven't experienced what the customer is actually going through in using that product" — whether the experience was frustration or joy.
He still does it. Tangen asked whether he wears Lululemon, whose board he sits on: "I do. I'm wearing Lululemon right now. I'm going to a Lululemon store later today. I drive General Motors cars." At Tesla he drove a car off the line every day.
He formalized it as a rule and told Musk about it. "I've got a 20% rule that's a little different than like a Google 20% rule where you can 20% of the time work on whatever you want. I told him that 20% of my time is going to be on the front lines because I want to experience what the customers experiencing and what our employees are experiencing."
His argument for talking to frontline staff is that they already have the answer: "Frontline employees can tell you exactly what's wrong with the product because they're hearing it from customers all day long."
And the answers converge, which is what makes it usable: "You don't typically get 500 different answers. You typically get like three or four of the same answers over and over again and it's very revealing."
His own version at a company he ran: "When I ran my own company, I called the switchboard every day to make sure it was picked up on ring number one."
Bonus Insights
Asked what people do not know about him, McNeill said music: "I have just an absolute admiration, joy, and appreciation of music." A musical mother nearly sent him to university to study it, and he said the base-eight counting music taught him turned out to help with advanced math and engineering.
Tangen asked whether he sees a production line as a symphony. "When I see robots welding, 300 robots welding a chassis, it does look like a symphony a bit to me."
His advice to young listeners is two kinds of mentor. "Grab a mentor as soon as you can." The first kind is vertical — "There's vertical mentors, people that are ahead of you in the journey" — who supply wisdom and experience.
The second kind is horizontal, and the instruction is specific: "Grab five or six first-time CEOs who are non-competitive in a business similar to yours size maybe growth." Meet quarterly under Chatham House rules and bring the biggest problem each of you has. "I wouldn't be where I am today without a half dozen mentors."
McNeill's bottom line is that the order of the five steps is the whole method: questioning, deleting and simplifying have to happen before speed, and speed before automation, because automation and AI applied to an unfixed process only make a company wrong faster and at greater expense.
Products, Companies & Tools Mentioned
Tesla (Where the five-step algorithm was built, the Model 3 automation failure happened, and gross margins ran at 24–28% against an industry at 10%)
Toyota (The source of the velocity-of-cash metric, and four days from aluminum to a finished car against Tesla's 15)
Apple (The roughly 30% gross margin Tesla set as its own target)
DoorDash (His example of automating last: PDFs of menus and a phone number before any software)
Lululemon (A board seat he says he uses as a customer, wearing the product and visiting stores)
General Motors (His example of founder-era culture surviving about 150 years, and a car he drives)
BYD (Named alongside Tesla as a vertically integrated Chinese manufacturer innovating the same way)
Lyft (Where he was COO after Tesla)
DVx Ventures (The firm he now runs, building companies)
Books & Resources Mentioned
The Algorithm – Jon McNeill (The book this interview walks through, step by step)
Made in America – Sam Walton (The book that convinced him to sample his own product every day)
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: