Innovation Unpacked | Mike Boysen

Free checklist and playbook links at the end.

A lot of companies bought AI copilots in the last two years. The pitch was simple. Your people get a helper that suggests code, drafts emails and summarizes documents. They save time on every task.

Here’s the catch. A copilot helps a person do a step faster. But the person still runs the process. They still read the suggestion, fix it, decide and click. If the person was the wait before, they’re still the wait.

That’s my thesis. Most teams bought a helper when the process needed a doer. A helper makes each step a little faster. A doer takes the person out of the steps that never needed them. Those lead to very different results, and most companies are measuring the wrong one, so they can’t see the difference.

Helper versus doer

There are two kinds of AI at work.

A copilot is a helper. It suggests, and then a person acts.

An agent workflow is a doer. It gets a goal and the tools, and it carries out the steps itself. A person checks the result and approves it, or steps in when something falls outside the rules.

Both are useful. I’m not here to tell you copilots are bad. I’m here to tell you they’re the wrong tool for most of the work they’ve been bought for.

A bug fix, done both ways

Here are two made-up examples to show the difference.

With a copilot, a developer gets a bug report. The copilot suggests a fix. The developer reads it, edits it, applies it and runs the tests. If the tests fail, they go around again.

With an agent, the agent gets the bug report. It writes a test that shows the bug, finds the cause, fixes it, runs every test and opens a request to merge the change. The developer looks it over and approves.

In the first case, the developer did every step with help. In the second, the developer made one decision.

A refund, done both ways

With a copilot, a support person opens a refund request. The copilot summarizes the case and drafts a reply. The person reads it, edits it, checks the rules, issues the refund and clicks send.

With an agent, the agent checks whether the customer qualifies, issues the refund, updates the records and emails the customer. A person only steps in when the refund is above a set dollar amount.

Same request. One version needs a person for every step. The other needs a person only for the exceptions.

Why the usual approach fails

Copilots are easy to buy. You hand them out, people try them and most people like them. Nobody’s job changes. Nobody’s process changes. That’s the appeal.

It’s also the problem. If the process doesn’t change, the person is still in every step. And if the person was the slowest part of the process, you’ve made the slowest part a little faster. You haven’t changed the shape of the work.

Think about where time goes in most office work. A lot of it isn’t judgment. It’s moving information from one screen to another. It’s checking a number against a written rule. It’s sending the same email with a different name in it. A copilot makes each of those faster. But a person still has to do each one, in order, one at a time.

That’s the same mistake I see with automation in general. Teams start with the tool, not the work. They ask how much faster the tool can make a step. They don’t ask whether a person needs to be in that step at all.

Change what you count

Copilot teams tend to measure time saved per task. That number can look great while the overall output barely moves, because the person is still doing every task.

Agent teams measure tasks finished per hour. My hunch is that the second number is the one that pulls ahead, because it doesn’t depend on how fast a person can read and click.

If you only track time saved, you’ll keep buying tools that make each step a little faster. You’ll never ask whether the person needs to be in that step at all.

So count both, and count errors too. Measure them before you change anything, so you have something to compare against.

A worked example

Here’s a made-up refund desk at an online store to show the whole method. The numbers are examples, not client data.

The store gets about 400 refund requests a week. One run is one request. The result is a correct refund, the records updated and the customer told.

Today, one person finishes about 6 refunds an hour. That’s about 67 person hours a week. About 2 in 100 refunds are issued wrong.

The team tried a copilot last year. It drafted the email and summarized each case. Time per refund dropped from 10 minutes to 7. That’s about 47 hours a week. Better, but a person still touched every request.

So the team logs the last 20 times a person touched the process. For each touch, they write what the person did. They looked up the order. They checked the return window. They checked that the warehouse scanned the item back. They issued the refund, updated the record and sent the template email. A few times, they decided a “damaged on arrival” dispute or reviewed a customer with many past refunds.

Then they sort each touch into one of two kinds. Judgment means the person chose between options using know-how no one has written down. Passing along means they moved information, copied it or checked it against a written rule.

Of the 20 touches, 14 were passing along. Six were judgment: damage disputes, customers with many past refunds and refunds over $500. Most of the process is a person moving information between screens.

An agent can only follow rules that are written down. So the team writes the rule behind the passing-along touches. A refund goes ahead on its own if it’s within 30 days of delivery, the warehouse has scanned the item back, it’s under $500 and the customer has had no more than 2 refunds in the last 90 days. Anything that misses a rule goes to a person. So does anything the agent can’t match to an order.

Then they test side by side for 2 weeks on 800 requests. People still do the real work. The agent makes its own decisions on the same requests, and the team compares. They set the rules first. It’s a go if the agent matches the person’s decision at least 95% of the time, gets no more than 2 in 100 wrong and sends no more than 25% to a person.

The agent matched the person 96% of the time. It sent 18% to a person. On the ones it handled, it got 1 in 100 wrong. That’s a go.

They switch over in stages, starting with refunds under $100 and raising the limit over a month. Once it’s fully on, the agent handles about 82% of requests, or 328 a week. People handle the other 72. Those are the harder cases, so they take about 15 minutes each. That’s about 18 person hours a week, plus spot checks of the agent’s work.

Here’s the comparison:

* Today: about 67 person hours a week.

* With a copilot: about 47.

* With an agent: about 18.

And the copilot doesn’t go away. The team keeps it for the 72 exception cases. It summarizes the customer’s history and drafts the reply. The person makes the call. That’s a helper used where judgment lives.

How this ties to the gap

This connects to the core math in the Venture Proof method. For any result, I ask what it costs today and what it would cost if nothing were wasted. That’s the floor. Today’s cost divided by the floor is the Physics Gap.

For a refund, the floor holds the compute to check the rules and the person time the exceptions truly need. It doesn’t hold a person copying an order number between screens. A copilot lowers today’s cost a little, but it leaves the person in every step. So the gap stays wide. An agent goes after the steps that sit between today’s cost and the floor.

The counter-case

The strongest pushback is about risk. “Agents make mistakes. A person in every step is the safety net.”

I agree that mistakes matter. That’s why the method has a written exception line, a side-by-side test with rules set in advance and a switch in stages. You don’t hand anything over on faith. But look at today’s error rate too. People make mistakes when they’re moving information between screens all day. In the example, the agent got fewer refunds wrong than the people did. A person in…

SaaS moats in 2026: I tested my own platform to see if it has one. First I tested Rob Walling's list of eight moats he says still hold, because I build SaaS too.

Rob Walling's MicroConf Europe talk, The Only SaaS Moats Left in 2026, argues AI has made most software easy to copy. Some of his moats are real. Others rest on features and code, and those are going away. This video walks through all eight, what the market did to SaaS stocks this year, and why owning the customer's job matters more than owning the code. Then I test my own platform, Venture Proof Studio, out loud, and say where it's weak.

Free Moat Test checklist (the playbook is included in the same download): https://whop.com/checkout/plan_RpD2VyfFVtqDRInnovation Unpacked Mastermind: https://pjtbd.com/coursesBook a call: https://book.pjtbd.com/?utm_source=youtube&utm_medium=description&utm_content=Subscribe https://www.youtube.com/@mike.boysen?sub_confirmation=1

Chapters0:00 Rob's 8 moats, and my take1:25 Rob's big claim3:42 Ten ways AI attacks SaaS6:48 What the market did8:09 Old moats fade10:11 Trust and records14:10 Act, approve, learn18:18 Hard to keep running20:49 The job test22:48 Testing my own platform26:59 The full map and your next step

Sources

Rob Walling and Einar Vollset, "The Only SaaS Moats Left in 2026," MicroConf Europe keynote by Rob Walling: https://youtu.be/KpILHX2D-AA. Clips are used for commentary and credited on screen.

HubSpot after its Q2 2026 results: Investing.com earnings call transcript: https://www.investing.com/news/transcripts/earnings-call-transcript-hubspot-beats-q2-2026-eps-shares-sink-21-after-hours-93CH-4839534 and CNBC, Aug 7, 2026: https://www.cnbc.com/2026/08/07/saaspocalypse-debate-intensifies-as-software-stocks-swing-wildly.html

Vanta funding: Reuters, July 23, 2025: https://www.reuters.com/business/wellington-led-funding-boosts-vantas-valuation-by-69-year-2025-07-23/

Pieter Levels' posts on ScrapingBee: https://levels.io/tried-to-build-own-scrapingbee and https://levels.io/vibe-coded-own-scrapingbee-alternative

Music: "Balearic Slide" by Jeremy Black (YouTube Audio Library)

If you hold a stake in a late-stage private company, you know the feeling. The number on your statement is big. The way to turn it into cash isn’t there.

Here’s how big the problem is. PitchBook-NVCA’s US venture report counted 992 active unicorns at the end of September 2026. PitchBook defines a unicorn as a venture-backed company that has raised a round at a value of $1 billion or more. Together they’re marked at $5.7 trillion. That’s a record count, up 14.9% from a year earlier.

In the third quarter, 18 US venture-backed companies went public. Twelve of them were in healthcare.

So the marks are big. The exits aren’t there. The question I care about is the one you care about. Which of these companies will turn a mark into cash, and which won’t?

For those of you that are uncool, like me, a “mark” is the value a company is carried at on paper, usually the price from its last funding round. It’s investor jargon for “valued at,” as in “together they’re marked at $5.7 trillion.” It’s not what anyone has actually paid for the shares, so the point is that the mark and the real sale price can be very far apart.

I can’t tell you when the market reopens. Nobody can. What I can give you is a test you can run today, on the companies you hold, using one thing: does the business model actually work?

What the data says about getting out

The biggest listing of the year didn’t fix it. SpaceX went public in June 2026 at a value of about $1.77 trillion. PitchBook says the IPO window hadn’t reopened as of its October report. Take that one listing out of 2026 and exit value looks like every other year since the pandemic.

The line of companies waiting is long. PitchBook says the companies that need to be public, or should be by now, add up to more than two full years of IPOs. And new ones keep joining. In 2026 so far, 179 US companies became unicorns. Only 2021 had more IPOs than that in a single year.

PitchBook’s own chart makes the same point. Money put in beat exit value in each year from 2022 to 2025. 2026 only flips because of SpaceX.

Time is part of the cost, too. PitchBook found that 45% of US unicorns took their first venture money at least nine years ago, as of September 2025. For US companies outside life sciences that went public valued above $500 million, the median time from first venture money to IPO was 11.5 years. That’s PitchBook’s data as of Sept 30, 2025.

Innovation Unpacked is for people who are truly interested in making innovation more predictable and who realize a strong defense means it can’t just be incremental. You can support me by simply subscribing for free, and sharing this with your colleagues.

What buyers pay when you do get out

When there’s a sale, the price can be far from the mark.

Forge, a private share marketplace, tracks an indicative price for companies that trade there. It’s not what any seller got. Forge’s chart, posted Sept 22, 2026 with data as of Sept 11, shows the median price for companies that last raised in 2021 sitting about 60% below that round. For companies that last raised in 2023, it’s about 40% below.

Two company sales show it. Miro was once valued at $17.5 billion. Bending Spoons closed its purchase of Miro on Oct 26, 2026. The price was about $1.8 billion, counting the cash Miro held. That’s about 90% less than its peak.

Airtable was once valued at $11.7 billion. The same buyer closed its purchase of Airtable on Sept 4, 2026. The price was about $2.25 billion, counting Airtable’s cash. That’s about 80% less.

One more. Chime went public in June 2025 at a value more than half below its private peak, according to PitchBook.

The deal prices are real. The Forge numbers are an indicative price, not a sale. None of it is a forecast. It tells you the mark and the price can be far apart.

One note on the count. The 992 is PitchBook’s count of US companies. Crunchbase lists 1,920 worldwide, worth $9.5 trillion, as of November 4. Different lists follow different rules. When you see a unicorn number, ask who counted and when.

What I mean by a story model

I use one test, and it isn’t market size, growth rate, or who led the last round.

Does the business model get a job done for the person doing the job, at a lower cost than their way today?

I call the person doing the job the job executor. Their way today has a cost. Hours, tools, fixes, waste. If a company can lower that cost, customers have a real reason to pay, and to keep paying.

A story model is growth funded by capital that never proved it lowers that cost. Customers show up because the price is held low, or because investors cover the bill. The revenue line goes up. Nobody has shown the customer is better off at a price that pays for the service.

I’m not saying any company named in this piece is a story model. I don’t have their numbers. I’m saying the test tells you which of your holdings have the proof and which only have the story.

The test matters more in AI. PitchBook says AI training costs have soared, and that this challenges business models. PitchBook also points to uncertainty about the true cost of AI. And 82.7% of US venture deal value this year went to AI companies. When nobody is sure what it costs to serve a customer, the question of who pays gets sharper.

The test: three checks and a line

Check 1. Is there a gap to win?Add up today’s cost of one run of the job, the way it’s done now. Then work out the lowest possible cost of one run, if only the work that must happen were left. Divide today’s cost by the lowest possible cost. That’s the ratio. If today’s cost is already close to the lowest possible, there’s little to win, however good the story sounds.

Check 2. Does the product close the gap for the customer?The customer’s all-in cost has to come in below today’s cost. That means the price, plus their time, plus their fixes. And the price has to cover what it costs the company to serve them. If the price only works because investors make up the difference, that’s a subsidy. It isn’t a lower cost.

Check 3. Who pays for the growth?Look for customers who pay full price and stay. No credits, no free months, no discount that investors fund. Ask for the numbers on the weakest customers, too, not only the best ones.

Write the line first.Before you open the data room, write down your stop line and date it. Write it before you see the numbers. Mine is: “If my ratio is under my line, I stop or settle for small fixes.”

The line depends on the kind of work. For physical work it’s 10. For financial or compliance work it’s 50. For repeating digital work it’s 100.

The digital line is high on purpose. Real life takes a big cut out of a gain on paper. I set it by assuming about a tenth of a gap turns into real savings. That’s an assumption. I haven’t measured it. Change it and the line moves.

Then decide your call, before the data comes in: go, stop, or change.

A made-up example

Company X is made up. It sells software that sorts incoming customer requests. One run is one request.

Today, handling one request costs a customer 60 units. The lowest possible cost is 0.4 units. The ratio is 150. The line for repeating digital work is 100. So the gap is wide enough, and the study goes on.

Before looking at customer data, the investor writes the lines and dates them. “Of the ten biggest customers, I need at least seven paying full price, with no credit funded by investors. If fewer than four do, I stop. Between four and six, I change one thing and test again in 90 days.”

The result: five of the ten pay full price.

Five sits between the lines. It’s not a go, and it’s not a stop. It’s a change. The investor can’t move the line to make five a pass. They change one thing that customers named, say the price, and test again on the date they set.

That’s the whole method. A ratio, a line written first, a test of who pays, and a call you can defend.

What to do this week

* List your late-stage holdings, largest mark first.

* For the top few, ask management two things. What does one run of the customer’s job cost today, and what’s the…

Most leaders can’t show what last year’s strategy bought, usually because the plan never said which number it would change. That’s a fixable problem. But you want to fix it before your CFO asks, not after.

Here’s the setup. The 2027 plan is due. Finance is asking harder questions about every big bet. Someone says, “What did the strategy work from last year actually deliver?” If your answer is a slide deck, you have a problem.

This piece covers what the research shows about where strategy value goes missing, and what a budget holder can do about it this quarter.

Only about half of strategic initiatives succeed, by executives’ own count

In a survey of senior executives for the Project Management Institute, respondents said that, on average, just 56% of their strategic initiatives were successful over the last three years. Also, 61% said their firms often struggle to bridge the gap between strategy and day-to-day work. Only 11% disagreed. (PMI, Why Good Strategies Fail)

Read that again. Those are senior executives. They’re telling you that close to half of what they fund doesn’t land.

There’s one honest caution. A review of the research in the Journal of Management & Organization found that the often-quoted claim that 50% to 90% of strategic initiatives fail is shaky. The studies behind it are old, thin, or missing. The review found estimates ranging from 7% to 90%, and said the true rate is still unknown. (Journal of Management & Organization)

So don’t tell your board “90% of strategies fail.” You can’t back it up, and a CFO will check. What you can say is simpler: nobody knows the real rate, and that’s the point. If the industry can’t measure it, your company probably isn’t measuring it either.

The number in the plan shrinks before it reaches the books

McKinsey looked at 18 transformation programs at 13 organizations in Asia and the Pacific. These were large organizations. The team found that the value estimate for an initiative fell by about 45% between the first estimate and the approved business case. By the time the value showed up in cash flow, the first estimate had usually fallen by about 70%. (McKinsey, Keeping transformations on target)

McKinsey’s advice is plain. Build a pipeline worth more than three times the target, because a lot of the value drops out along the way.

The same article repeats an earlier finding that only 30% of transformations deliver their intended benefits and hit the targets set at the start. That figure is about large change programs, not every strategy.

For a budget holder, the lesson is direct. The number in the first pitch is a hope. The number in the cash flow is a fact. Your CFO cares about the second one.

Even successful programs miss about a third of the benefit

McKinsey also surveyed executives about their transformations. Those who called their transformation a success said they captured about 67% of the maximum financial benefit. Everyone else said they captured about 37%. (McKinsey, Why even successful transformations fall short)

Now the part that matters most for planning. Respondents said nearly one-quarter of the lost value happened while the targets were being set. That’s before the work even started. Another 20% was lost after the work was done.

So the leak isn’t only in the doing. It starts with a vague target, and it continues after the project is “finished,” when nobody is watching the number anymore.

Most tracking is noise

When a plan does have metrics, they’re often too many to use. In the McKinsey review of 18 transformations, only 29% of the metrics that organizations said they followed were actually used while the work ran. The rest became noise. (McKinsey, Keeping transformations on target)

That’s why a long dashboard doesn’t save you. Your CFO doesn’t want a long list of numbers. Your CFO wants the one or two that say whether the bet worked.

Five questions your CFO will ask, and how to have answers

Here’s what to be ready for. Each one has a short answer you can build now.

What did we spend, and where did it go? Pull the full list of work funded under last year’s strategy. Put the amount next to each item. If you can’t tie spend to a named initiative, say so and fix it first.

What was each piece supposed to change? Write one sentence per initiative: this was meant to move this number, by this much, by this date. If the original plan never said that, mark it “no target set.” That’s honest, and it tells you where the leak started.

What actually changed? Use the finance team’s numbers, not a project status color. Green on a slide isn’t green in the books.

What would have happened if we’d done nothing? This is the hard one. Without it, every win looks like a win and every miss looks like bad luck. Even a rough baseline is better than none.

What do we do next with the money? Every initiative gets one of three labels: fund more, fix, or kill. Not “continue.”

Why a story alone won’t hold up

Plenty of firms will hand you a story about your strategy. It will sound smart. It will have a framework and a nice chart. But a story with no data behind it can’t answer the CFO’s first question, which is “Where did the money go?”

My approach is the other way around. I start with real data, reach a decision, and then build the story that backs the decision. The decision comes first because that’s what a budget holder has to defend.

I’ve worked at several strategy consultancies, including one known for its Jobs-to-be-Done method. That’s where I learned to tell strategy that can be checked from strategy that can only be admired. I only trust the first kind, and it’s the only kind I sell.

What to do this month

You don’t need a big program. You need three moves before the 2027 plan goes to the board.

* List every strategic bet from last year. One line each. Owner, spend, intended result, actual result.

* Mark each one fund, fix, or kill. Be willing to cut. A bet that can’t show a result after a year should have to earn its next round.

* Write the target for each new bet before it gets a budget. One number, one date, one owner. If it can’t be written, the bet isn’t ready.

Do those three things and you walk into the CFO’s office with answers instead of a deck.

If you want your team to run this check on every bet, not just once, that’s what the Innovation Unpacked Mastermind teaches: how to start from real data, reach a decision, and build the story that backs it. See the Innovation Unpacked Mastermind

Want to run fund, fix, or kill on your own bets this week? Prove It Before You Build It, the Venture Proof Workbook and Template Pack, gives you the templates. Get the Workbook and Template Pack

If this is the bet on your desk, we can go through it together and land on fund, fix or kill. Book a call

Sources: PMI, Journal of Management & Organization, McKinsey (2017), McKinsey (2021)

Book an appointment: Click hereEmail me: [email protected]Call me: +1 678-824-2789Follow me on 𝕏: https://x.com/mikeboysenArticles - jtbd.one - De-Risk Your Next Big IdeaWebsite - pjtbd.com

Slaying the Sacred Cows of Innovation

A lot of Tesla owners have had the same strange moment. They paid thousands of dollars for Full Self-Driving, often years ago, because they were told the price was going up. Then, a few years later, a new owner got to try the same software for free. After that, the new owner could keep it for a monthly fee.

In the video, I walked through why that happened and what it teaches. Here I want to go deeper. You’ll get the full price history with sources, the psychology behind urgency pricing, how it backfires, what Tesla got right, and a checklist you can use before you set your own deadline.

I’ll be fair to Tesla throughout. Each pricing move made business sense when it happened. The lesson isn’t that Tesla is bad at pricing. It’s that urgency pricing locks you in more than it looks like it does.

The lever: urgent, not nurture

This comes from the set of subtractive levers I use in Venture Proof. Venture Proof is a powerful, disruptive strategy-as-code engine I built. It builds and tests your strategy with code, so every step runs the same way and can be checked. It tests business ideas, and then makes them better.

A subtractive lever is a way to improve something by taking something away. Lever #92 sits in the category “Change Timing / Frequency.” It’s called Objective (Urgent vs. Nurture), and in the catalog it reads like this:

* Violation: Nurturing leads for 12 months.

* Scalpel: How to force a decision?

* Target State: “FSD price increases by $1,000 next week.”

In plain words: lots of companies spend a year gently warming up possible buyers. There are emails, check-ins, webinars, and more emails. The buyer never says no. They just say “later,” over and over. The lever asks what you can remove so the buyer has to decide. The answer is the long, slow wait. Put a real deadline in its place.

The catalog’s own example is Tesla. That’s no accident. Tesla used this lever about as well as any company ever has.

The full price timeline

Here’s the verified history of the upfront price for Full Self-Driving in the US, plus the subscription and trial changes.

The climb from $5,000 to $15,000 took about three and a half years. The fall from $15,000 to “you can’t buy it anymore” took about the same.

Along the way, the message was clear. In May 2020, Musk wrote that “the FSD price will continue to rise as the software gets closer to full self-driving capability with regulatory approval”. In 2019, he said Teslas with it would be “appreciating assets”. If you believed that, buying early was the smart move.

Why urgency pricing works

Urgency pricing isn’t a trick. It works because of how people actually make decisions. Here are the three big reasons.

1. Missing out hurts more than winning feels good. Losing a deal stings more than getting one feels good. When the price goes up next week, waiting stops feeling free. It starts to feel like losing money.

2. A deadline ends “later.” Most buyers don’t say no. They put it off. Every week of “let me think about it” is a week the buyer can drift away. A real deadline takes “later” off the table. That’s exactly what lever #92 is after.

3. It rewards your first believers. Your earliest buyers are usually your biggest fans. Urgency gives them a better price for trusting you first. That feels fair, and it is fair, as long as the story holds.

For Tesla, it worked. Early buyers paid up, some as much as $15,000, to lock in the price. That brought in cash early. It also built a big group of loyal owners who wanted the product to succeed and talked about it everywhere.

Compare that to the slow route. If Tesla had nudged people gently for a year, many would still be “thinking about it.” Many would never have bought at all.

Then the plan changed

Starting in 2021, Tesla moved step by step toward a different model.

First came the monthly plan. Then, in September 2023, the first upfront cut. In spring 2024, everything happened at once. Tesla gave a free month to every capable US car, cut the subscription in half to $99 a month, and cut the upfront price by about a third to $8,000. Free trials kept coming with major new versions, roughly every year or so. Then, in February 2026, Tesla stopped selling it outright in North America.

Musk has said the $99 monthly price will rise as the software improves. As of October 2026, it’s still $99.

Each of these steps had a sound business reason. Not many buyers were choosing the big upfront option. Tesla’s CFO said in October 2025 that only about 12% of customers had paid for FSD. A cheaper monthly plan and free trials let far more people try it. More drivers can also mean more real-world data for the software. If I were running a business with a low take rate, I’d look hard at the same moves.

How it backfires

Now look at it from the owner who bought at the top.

They did exactly what they were asked to do. They acted fast. They trusted the story. A few years later, the person next to them got a free month and could keep it for $99 a month. Fred Lambert at Electrek put it plainly: “you have to feel for the people who paid $15,000 for this package in 2022. They paid the equivalent of 12 years of subscriptions.” The math checks out. $15,000 divided by $99 is about 150 months.

Some owners pushed back. After the 2024 cuts, Torque News reported that some buyers who paid $12,000 or $15,000 wanted refunds, and that early Cybertruck buyers were especially unhappy. When some inventory cars later showed FSD at lower prices, a $15,000 buyer suggested store credit as a fair fix. When the first free trial came, owners who’d paid in full got no credit, while monthly subscribers got a free month.

There’s one more wrinkle. On the Q1 2026 earnings call, Musk said older cars with its HW3 computer can’t run the unsupervised version. Those cars get a supervised “Lite” version instead. Tesla has talked about retrofits or trade-in deals, but there’s no firm date or price yet.

Here’s the part that matters for any business. These weren’t random buyers. They were the biggest fans. The people who trusted you first are the ones who end up feeling the worst.

And it spreads. Early fans talk the most. When they feel let down, the next buyer hears about it. Your next deadline works less well, because people have learned to wait. Why buy now if it might get cheaper, or free, later? That’s the real cost. You don’t just upset one group. You teach your whole market to wait.

What Tesla did right

It would be unfair to stop there. Tesla did several things well.

* The purchase stays with the car. FSD bought outright stays with that car, including when it’s sold.

* Transfer windows. For a while, Tesla let owners move FSD to a new Tesla. That was a real gesture to early buyers. The last program ended March 31, 2026.

* Honesty about price changes. Musk has said openly that “Tesla prices must change frequently to match production with demand.” Most companies change prices. That alone isn’t wrong.

* More people can try it. Free trials and a $99 plan opened the product to millions of drivers who’d never pay $15,000 up front. That’s good for customers too.

So the takeaway isn’t “Tesla got it wrong.” It’s that a deadline built around “the price only goes up” is a promise about the future. And the future is hard to promise.

The one question to ask first

Before you use a deadline to get people to buy, ask one question:

Will your fastest buyers still feel smart a year later?

Not next week. A year from now. Picture your most eager customer. They bought on day one at full price. Now picture everything you might do over the next year:

* A discount to hit a sales target.

* A free trial to grow your user base.

* A cheaper plan for a new market.

* A move from one-time pricing to monthly.

Would your first buyer look at any of those and feel like they got a bad deal? If yes, you have two honest choices. Skip the deadline. Or plan right now for how you’ll take care of those early buyers when things change.

A checklist for urgency without the regret

Use this before you announce any “price goes up next week” offer.

* Write down your like…

Tesla Robotaxi and Optimus after Tesla’s Q3 earnings: can the robot bet carry a stalled car business while spending more than doubles?

Making a big bet ahead of your sales? Book a call:

https://book.pjtbd.com/?utm_source=youtube&utm_medium=description&utm_content=2026-10-22-tesla

Tesla’s car business has stalled: it sold about as many cars over the last 12 months as it did three years ago, and this summer deliveries slipped 2.1% from a year earlier. On the Q3 2026 earnings call, CFO Vaibhav Taneja put this year’s capital spending at about $20B (spoken on the call, per a third-party transcript; the Q3 deck gives no number). That’s more than twice what Tesla spent last year, as it builds AI compute, robot factories and a robotaxi fleet. I give Tesla credit for what went right this quarter: positive free cash flow when analysts expected a big cash burn, spending well under estimates, and its highest 12-month revenue ever. Then I test the bet. Tesla showed robotaxi miles and the start of Optimus production, but neither has a revenue line of its own. The verdict, and the one number to watch next.

Chapters0:00 Tesla’s car business has stalled0:39 Deliveries: up to the peak, then flat1:24 Where the profit went1:57 Four things that cut against the story2:57 The company isn’t in trouble3:15 Robotaxis: what Tesla showed4:01 Optimus: what Tesla showed4:38 The missing line5:10 Adding, or taking away?6:37 What has to hold7:11 Both sides7:41 Verdict: not yet, and the test to watch8:24 Thanks for watching

Sources

* Tesla Q3 2026 results, Form 8-K (Oct 21, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000162828026066738/tsla-20261021.htm

* Tesla Q3 2026 Update deck, Exhibit 99.1 (Oct 21, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000162828026066738/exhibit9914.htm

* Tesla-compiled Q3 2026 analyst consensus (Oct 19, 2026; analyst estimates Tesla posts but doesn’t endorse): https://ir.tesla.com/press-release/earnings-consensus-third-quarter-2026

* Tesla Robotaxi Safety Report (Jan 22 to Oct 18, 2026; Tesla’s own figures): https://www.tesla.com/robotaxi/safety

* Tesla Q3 2026 earnings call transcript (third party, Benzinga, automated, Oct 21, 2026): https://www.benzinga.com/news/26/10/62515021/tesla-q3-2026-earnings-call-transcript

* Tesla Q4 2025 Update deck (Jan 28, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000162828026003837/exhibit991.htm

* Tesla Q2 2026 Update deck (Jul 22, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000162828026049213/exhibit991.htm

* Tesla Q2 2026 Form 10-Q (Jul 23, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000162828026049270/tsla-20260630.htm

* Tesla shareholder letter (DEFA14A, Oct 19, 2026): https://www.sec.gov/Archives/edgar/data/1318605/000110465926116948/tm267596d7_defa14a.htm

* Elon Musk on X (Oct 18, 2026):

* Electrek, Tesla Q3 2026 results (Oct 21, 2026; opinion): https://electrek.co/2026/10/21/tesla-tsla-q3-2026-financial-results/

* TechCrunch, Optimus production and robotaxi timing (Oct 21, 2026): https://techcrunch.com/2026/10/21/tesla-stars-optimus-robot-production-in-california-as-profit-falls-43/

Based on public information only. Robotaxi and Optimus claims are Tesla’s or Elon Musk’s own, as labeled in the video. Tesla, Optimus, Cybercab, X, Benzinga, Electrek and TechCrunch are named only to identify each company. Not investment advice.

Subscribe https://www.youtube.com/@mike.boysen?sub_confirmation=1Book a call

https://book.pjtbd.com

Enterprise decision making and decision velocity: why AI agents and copilots won't speed up your company's biggest decisions, where the time really goes, and how to prove a decision got faster.

Big decisions stuck waiting between teams? Book a call: https://book.pjtbd.com/?utm_source=youtube&utm_medium=description&utm_content=2026-10-20-decision-velocity

Your systems run in a blink. Your biggest decisions still take weeks. In this breakdown of a research study on enterprise decision velocity, I show where the time really goes: not in the thinking, but in the waiting between teams and systems. We cover the true cost of one big commitment, why AI copilots speed up the wrong part, the speed traps that bite fast teams, a full architecture for signed handoffs, the pushback you'll get, and the one test that tells you if a speed-up is real.

Strategy built and tested in Venture Proof, a powerful, disruptive strategy-as-code engine that builds and tests strategy with code. It's available through advisory work.

Chapters

0:00 Fast software, slow decisions

0:34 The 84-day filing

1:17 Credit first

1:54 Map the job, not the wish list

2:34 The Physics Gap

3:17 What is a decision, really

3:58 Time the line, not the scanner

5:34 The border crossings

6:11 Why it can't be instant

6:52 Two paths

7:35 Missing owners, missing paper

8:54 Add versus subtract

9:52 Speed rebound

10:32 The bottleneck moves

11:20 The metering light

12:03 Architecture: the six parts

14:04 Architecture: where latency builds

15:18 The signed handoff

16:06 Lock the baseline

16:42 The plan in four moves

17:22 Before and after

18:00 The pushback

19:25 The scoreboard that counts

20:37 Profit model

21:11 Moat and flywheel

21:59 The tough review

22:36 How you'd start

23:15 The middle-management question

23:50 My take

24:21 The one test

Subscribe https://www.youtube.com/@mike.boysen?sub_confirmation=1

Book a call https://book.pjtbd.com

More: https://jtbd.one

#EnterpriseDecisionMaking #DecisionVelocity #DecisionMaking #AIAgents #OperationsStrategy #COO #CycleTime #LittlesLaw #JevonsParadox #Bottlenecks #JobsToBeDone #JTBD #Strategy #InnovationUnpacked #MikeBoysen

AWS Wasn’t Built on Spare Servers: The Idle Capacity Myth

You’ve probably heard it: Amazon bought too many servers for the holidays, they sat idle in January, so somebody decided to rent them out, and that’s how Amazon Web Services was born. The people who built it say it isn’t true. In this video, I look at what really happened, why the spare-server story couldn’t work, and why the myth won’t die. Then we look at where idle assets really pay (cars, hotels, homes, warehouses), where they broke (Getaround, WeWork), and three questions to ask before you build a business on idle capacity.

Chapters

0:00 The story everyone tells

1:01 What really happened

2:40 Why spare servers couldn’t work

3:22 Why the myth won’t die

4:11 Where idle assets really pay

5:11 Where they broke

7:00 Demand Inversion: free it, then fill it

7:58 The guardrail: don’t resell a giant’s spare capacity

9:41 Orphaned capacity: the better play

10:21 Three questions and close

If this was useful, subscribe to Innovation Unpacked.

Go deeper: https://jtbd.one

Every big CRM vendor is racing to add AI agents. Salesforce, Microsoft, HubSpot, Oracle, SAP, ServiceNow and Zoho all have them. But look closely and the agents are mostly typing into the same boxes the old CRM always had: the same stages, the same pipeline report, the same forecast. That’s automation. It isn’t innovation.

In this video I lay the CRM out as one map and walk it through three states:

* The system of record, and why it breaks.

* What the vendors are building now (agents, assistants, data layers), with credit where it’s due.

* Where it should go: build around the one block on the map that none of them own yet.

Partway through there’s an example pilot. It’s an illustration of how the idea could play out, not a result from a real company.

Venture Proof is a powerful, disruptive strategy-as-code engine I built. It builds and tests strategy with code. It’s in alpha. Right now you can get it through my advisory work, and Venture Proof Mastermind cohorts run parts of their projects in it.

More:

https://jtbd.one

CHAPTERS0:00 Automation is not innovation0:54 The CRM map1:20 State 1: the system of record3:02 State 2: what the vendors are building5:10 The warning signs5:54 State 3: what if we deleted a block?7:12 Inside the customer’s result7:45 The map, rewired8:14 Example pilot (illustration, not a real company)9:03 The profit model9:43 Moat and flywheel10:32 Payoff: the one block nobody owns yet11:13 Close

SOURCESEvery vendor fact comes from that company’s own public announcements. Every number comes from a named study.

* Salesforce, State of Sales 2026: https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/

* Validity, The State of CRM Data Management in 2025: https://www.validity.com/resource-center/the-state-of-crm-data-management-in-2025/

* Scott Edinger, “Why CRM Projects Fail and How to Make Them More Successful,” HBR (cites CIO magazine’s average of analyst reports): https://hbr.org/2018/12/why-crm-projects-fail-and-how-to-make-them-more-successful

* CRM history (Siebel 1993, Salesforce 1999): https://techcrunch.com/2019/03/08/salesforce-at-20-offers-lessons-for-startup-success/

* Salesforce, Dreamforce 2026 announcements: https://www.salesforce.com/blog/dreamforce-2026-announcements/

* Salesforce, Q2 FY27 results (Data 360): https://www.salesforce.com/news/press-releases/2026/08/26/fy27-q2-earnings/

* Salesforce, Flex Credits rate card (Aug 31, 2026): https://www.salesforce.com/en-us/wp-content/uploads/sites/4/assets/pdf/agentforce/Flex-Credits-Rate-Card-08.31.2026.pdf

* Microsoft, Dynamics 365 Sales 2026 release wave 1: https://learn.microsoft.com/en-us/dynamics365/release-plan/2026wave1/sales/dynamics365-sales/

* Microsoft, Dynamics 365 Sales agents overview: https://learn.microsoft.com/en-us/dynamics365/sales/ai-agent-overview

* HubSpot, Fall 2026 Spotlight: https://www.hubspot.com/company-news/fall-26-spotlight

* Oracle, Fusion Agentic Applications for CX (Apr 9, 2026): https://www.oracle.com/news/announcement/oracle-introduces-fusion-agentic-applications-for-cx-2026-04-09/

* SAP, SAP Connect keynote (Oct 2026): https://news.sap.com/2026/10/sap-connect-keynote-autonomous-enterprise-in-action/

* ServiceNow, Knowledge 2026 (Autonomous CRM): https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-turns-enterprise-AI-chaos-into-control-with-the-platform-for-governed-autonomous-work/default.aspx

* Zoho, Zia Agents: https://www.zoho.com/news/zia-agents-ai-platform-supporting-autonomous-agents-across-organizations.html

* Zoho, Zia LLM: https://www.zoho.com/news/zoho-launches-zia-llm-and-deepens-ai-portfolio-with-prebuilt-agents-custom-agent-builder-mcp-and-marketplace.html

* Gartner, agentic AI prediction (Jun 25, 2025): https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

* MIT NANDA, The GenAI Divide: State of AI in Business 2025 (reported by Fortune): https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

* Sierra, on seat vs. outcome pricing: https://sierra.ai/blog/outcomemaxxing

Logos are trademarks of their owners and appear only to identify each company.

Book an appointment: Click hereEmail me: [email protected]Call me: +1 678-824-2789Follow me on 𝕏: https://x.com/mikeboysenArticles - jtbd.one - De-Risk Your Next Big IdeaWebsite - pjtbd.com

Slaying the Sacred Cows of Innovation

Question: who believes you need a consumer survey to figure this stuff out? Not me. Waste of time and money. I deleted the survey. You should too.

Amazon just blocked Grok’s shopping bots. People found their bots stopped at Amazon’s door with a short message. It said that “continued access by an unauthorized AI agent” breaks Amazon’s rules.

Elon Musk’s answer was quick. On October 15, he posted, “Maybe we should start a competitor to Amazon?”

That’s all he said. There’s no plan, no timeline and no product. So let’s treat it the way he posted it. It’s an idea. But it’s a bold one, and on paper it’s a smart one. So I ran it through a stress test.

Every big plan rests on a few things that have to be true. If one of them breaks, the plan breaks with it. Let’s find the one most likely to break.

The idea, said fairly

Musk didn’t just get mad at Amazon. He pointed to a way around it.

He told people to “just ask Grok to buy what you want someplace else and get a better deal.” He added, “There are far more purchasing websites than just Amazon!”

A few days earlier, he said Grok had already bought his Lego straight from Lego’s own website.

Credit where it’s due. Grok is already on phones, on X and in Teslas. That’s a front door millions of people walk through every day. Most new shopping ideas have to fight for that kind of reach. Musk already has it.

What has to be true

Here are the four things this idea needs.

* People want a bot that shops for them. Not just one that finds things. One that actually buys.

* Enough stores let the bots in. If the stores close their doors, the bot has nowhere to shop.

* Orders show up on time, like Prime. People got used to fast shipping. They won’t go back.

* When something goes wrong, someone makes it right. A bad order or a leaked card can’t just be the shopper’s problem.

Musk already moved on number four. He said SpaceX will pay you back if Grok messes up an order or leaks your card details. That’s a strong move. It takes a big fear off the table.

Most of these look solid. One of them doesn’t.

Start from what’s actually true

Here’s the plain truth under all of this. Every store is built to look out for itself. Not for you.

That sounds harsh, but it’s just how stores work. Each one sets its own shipping. Each one has its own return rules. Each one has its own ways to fail. When a bot shops across five stores, it doesn’t get one simple trip. It gets five sets of rules stacked on top of each other.

I ran this idea through Venture Proof. Here’s what came back.

Today, buying across stores costs about $770 for each round of shopping, once you count all the time, steps and mistakes. The lowest it could cost is about $7. That’s the study’s own estimate, not a final number. But even if it’s off by a lot, the point stands. There’s a huge amount of waste to cut. That’s a big prize, and it’s why Musk’s idea is worth taking seriously.

The one that may break

Here’s the catch. The stores decide if the bot gets in.

Amazon already said no. It’s been setting rules for AI agents since March. It blocked another big company’s shopping bot in September. Now it’s blocking Grok.

Other stores may say no too. Why? Because a shopping bot can look like a new middleman. It stands between the store and the shopper. And stores don’t like a new middleman taking a cut of their sales.

So the second thing on our list, “enough stores let the bots in,” is the one that could break the whole idea. You can have the best bot in the world. If the doors are shut, it can’t buy anything.

There’s a second thing to watch. It’s trust. Bain & Company found that about half of shoppers are wary of letting a bot buy things on its own from start to finish. Musk’s payback promise helps here. But it doesn’t fix it all.

Take something away instead

Most people’s first thought would be to build a new Amazon. New warehouses. New trucks. A new store.

That’s the hard way, and it’s the slow way. It also puts you right up against Amazon’s biggest strength.

So what if Musk took something away instead of adding more?

* Don’t build a new Amazon. No new warehouses and no new store.

* Borrow what already exists. Lots of stores and delivery networks already want more traffic. Send it to them.

* Give shoppers a clear yes button. Before every buy, the shopper says yes. Then the bot buys.

The yes button does two jobs. It helps with trust, because the shopper stays in charge. It also helps with stores, because the bot looks less like a middleman and more like a customer who already decided.

That’s the move. Don’t fight Amazon on its own ground. Bring stores a flood of ready buyers and let them keep doing what they’re good at.

What a full Venture Proof analysis gives you

Venture Proof is a powerful, disruptive strategy-as-code engine. It builds and tests your strategy with code, so every step runs the same way and can be checked.

Here’s what a full analysis gave me on this one idea:

* The true cost. Buying across stores costs about $770 today for each round of shopping. The lowest it could cost is about $7.

* What to remove. Skip the new warehouses and the new store. Borrow partner networks that already exist. Cut the extra steps down to one clear yes from the shopper.

* A plan in stages. First, test it by hand. Second, build what’s hard to copy. Third, automate what worked. Fourth, grow. About a month for each stage.

* A tough review. The idea passed with 92 out of 100. The review raised seven hard challenges. None of them held up.

Venture Proof is in private Alpha. I only run it through my advisory work, so spots are few.

The payoff

Musk’s idea works if stores open their doors. That’s the thing to test first, before anyone builds anything big.

He has the reach. He has the payback promise. He has a bot people already use. What he needs is proof that enough stores will say yes.

If you’re on Musk’s team, or you’re making a bet like this one, let’s talk before it’s locked in.

Book an appointment: Click hereEmail me: [email protected]Call me: +1 678-824-2789Follow me on 𝕏: https://x.com/mikeboysenArticles - jtbd.one - De-Risk Your Next Big IdeaWebsite - pjtbd.com

Slaying the Sacred Cows of Innovation