"A little rough" is not a plan
If you listen to the people building artificial intelligence long enough, you will hear the same phrase come up. The transition is going to be a little uncomfortable. A little rough. Something we should prepare for. Then the conversation moves on to the exciting part, the cures, the cleaner oceans, the abundance on the other side.
I want the abundance too. I build with these tools every day. But every time I hear "a little rough," I wait for the next sentence, the one that says what the plan is. Who it affects. How long it lasts. Who pays for it. That sentence rarely comes.
What we usually get instead is advice. Level up. Learn something new. Maybe learn about AI. That is fine as far as it goes, but it puts the entire weight of the transition on the person with the least power in the room. The worker is told to adapt. The company that made the decision is told nothing.
I think part of the reason the phrase gets dropped in so casually is that it works as a receipt. Later, when whatever "a little rough" turns out to be actually shows up, somebody can say, we told you. Maybe there is some guilt riding along with it. I do not know what they are seeing coming. I do know their pain will not be equal to the pain of the people standing on the ground floor looking up.
So this article is about the plan. Not a protest. Not a prediction that AI is going to destroy us. A practical idea for how we get through the middle of this thing with people still standing, and what each of us can do about it right now. I call the first piece the seat guarantee.
One ask: let us keep our jobs
This week I put up a website with one ask behind it. Let us keep our jobs. Keep the employment rate low while the technology does its work. You can read it at transitionmoonshot.com.
I believe that ask is achievable. The money is there. The productivity gains are real. The technology works. History says the technology almost always works. The part that goes wrong is what happens to the people in the middle while the gains get sorted out, and who gets to keep them.
There is one thing that gets in the way, and it is not a technical problem. It is greed. Not greed as a cartoon villain, but greed as a default setting. If nobody builds a different default, the easiest decision wins, and the easiest decision is to cut the person and keep the savings.
I want to be careful here, because this is where people expect me to start blaming AI. I am not going to. AI is not the thing that hurts people in this story. People using AI as a tool against other people is the thing that hurts people. That distinction matters, because you cannot negotiate with a technology, but you can absolutely set rules for the people using it.
If you want the bigger picture of why I started this project, start with the Transition Moonshot and then who pays the gap. This piece goes deeper on one mechanism.
A lease that goes with the job
Here is the idea I came up with the morning I recorded this episode. It comes straight out of real estate, which is the other half of my working life.
When a building sells, the tenants with a lease do not get thrown out the next morning. The new owner buys the building with the lease attached. There are rules in place. The tenant gets to finish out what they were promised, and the new owner plans around that from the start. Nobody thinks this is radical. It is just how property changes hands in an orderly way.
Now apply that to work. When AI takes a job, the person comes with the deal for a while. The company that captures the savings from the AI seat keeps paying the person through the transition, until they land somewhere new. That is the seat guarantee. It is the first part of the Transition Moonshot plan.
Notice what this does and does not ask for. It does not ask a company to stop using AI. It does not ask anybody to pretend the AI is not better at the task. It does not freeze anything in place. It asks that the handoff come with a bridge, paid for out of the very savings the handoff creates.
Does the company pay forever? No. A lease has a term. The seat guarantee is a period of time, long enough for a person to land, not a lifetime pension. I would much rather argue about the length of the bridge than about whether there should be one.
If you want the housing version of this analogy laid out with the tract, the HOA bill and the flip, it lives in The AI Boom Is A Housing Tract.
The seat math
Can a company afford it? Here is the arithmetic that makes me think yes.
A full-time person works roughly 1,880 hours a year once vacation and holidays come out. An AI system can run every hour of the year. That is 8,760 hours. Divide one by the other and an AI seat that is only as good as the person it replaces still delivers about 4.66 times the working hours.
Read that again, because it is the whole case. I am not assuming the AI is a genius. I am assuming it is merely as productive as one human per hour. Even at that level, the seat runs nights, weekends and holidays. It does not get sick. It does not need overtime. It does not burn out. It covers three shifts.
And that is the floor. If the AI is better than the person at the task, faster or more accurate or able to run as many copies as you need, the gain stops being a multiple and starts looking exponential. A company is not adding one agent working 8,760 hours. It may be adding a thousand.
So when someone says paying a displaced worker through a transition would break the business, I want to see their math. If the capacity gain is anywhere near this size, keeping a person paid for a defined period during the handoff is a rounding error against what the seat produces. The money is there. The only question is whether anyone decides to spend a small piece of it on the people who made the business work in the first place.
Treat these numbers as my framing, not a forecast for any one company. Every business has its own costs, and some jobs will not translate cleanly into hours at all. But the direction is the point. The surplus is big enough to carry a bridge.
Where the pressure comes from
In the episode I asked viewers to correct me on something, and I want to put the better answer in writing. My worry was that publicly traded companies are somehow required to cut people whenever AI is cheaper, that the rules force their hand.
As best I understand it, no law requires a public company to automate a job. Boards and executives answer to shareholders, and they get a lot of room in how they run the business. What creates the pressure is competition and the stock price. If a competitor cuts and reports bigger margins, the market rewards them, and every other board gets asked why they are not doing the same.
That is good news, in a way. A statute is hard to change. Pressure is easier, because pressure responds to incentives. If retaining people through a transition earns a company recognition, customers, talent and a story investors like, the pressure can point the other direction. That is the entire reason I think of this as a Moonshot problem. Prizes and public standards are how you move incentives without waiting for a law.
It also means companies that are not publicly traded may be the first ones able to lead. A private owner can keep their people through the change without explaining it to a quarterly call, and show everyone else it works.
There is a second pressure worth naming. When a company announces it is embracing AI, its valuation can go up, because it looks forward thinking. That creates a temptation to call any layoff an AI layoff, even when the real reason is something else, like having hired too many people a few years ago. I covered that problem, and the reporting checkbox that is supposed to catch it, in The Machine Didn't Sign The Layoff. The short version: if we do not get accurate attribution of why people lose their jobs, we cannot even measure whether a seat guarantee is working.
The quiet loophole: just stop hiring
Here is the hardest objection to my own plan, and I want to face it directly instead of hoping nobody notices.
If a company owes people something when it replaces them, the cheapest move is to never hire them in the first place. Nobody gets fired. There is no layoff headline. There is no one to guarantee. The work just quietly gets absorbed by AI and attrition, and the next person who would have been hired never gets the seat.
I believe we are already seeing some of this. From where I sit, hiring looks slower. People seem to have paused. The open roles that used to be the ladder into a career feel thinner. That is a much harder harm to see than a layoff, because there is no one to interview about the job they never got.
So does the plan hold or break here? For now I think it holds, with an amendment. The answer is to attach the obligation to the AI seat, not to the person. Call it a productivity dividend. If a company runs AI seats doing work that people used to do, a small share of that seat's value flows into the transition, whether the company fired somebody or never hired them. That closes the loophole, because it no longer matters which door the company walked through. The company pays on the seat either way.
This also covers the one-person company that never hires anyone and runs entirely on agents, which is coming faster than most people think. More on that below.
Who ends up paying for the AI seat?
A dividend on the AI seat sounds clean on paper, but I want to follow the money one more step, because I think it rolls downhill.
The big corporations will get enterprise AI. The best models, the custom deals, and a massive volume break because they are spending so much. The rest of us are on the $20 or $200 monthly plans for ChatGPT or Claude. If a cost gets attached to AI seats, I worry it lands hardest on the small users who have no negotiating power, the same people at the bottom of the elevator shaft looking up.
So any version of this has to be designed with the small operator in mind. A dividend that a giant enterprise shrugs off and a one-person shop cannot afford is not a transition plan. It is a moat.
Here is how I use AI myself, and why I think the small operator is worth protecting. I use it like a hammer. I have a workflow I need followed. I need it done a certain way. I need it to check back with me before anything ships. That is a great use of AI. It frees me up to do other work, and it keeps me responsible for where the hammer lands. Millions of small businesses are going to use AI exactly this way, and a transition plan should make that path easier, not more expensive.
The same problem from a different angle is in The AI Access Gap, Explained Simply.
Stanley and Stanley AI
I tell a story to make the transition feel real. Stanley is a hypothetical. He is not a real person and this is not a report about any company. But every piece of it is something a company could do this year.
Stanley has been with his company for 20 years. He runs a department. He is a team player. Everybody likes Stanley. One day an outside advisor convinces leadership they need to move the whole operation onto AI, but slowly, so nobody panics.
So they call Stanley in. They tell him AI is going to start helping with his job. Treat it like a counterpart. Teach it what you do. It will take work off your plate so you can focus on bigger things. And as icing on the cake, no performance evaluation for the next year while you get used to it. Stanley is thrilled. He is busy, and this sounds like relief.
Stanley gives his AI access to everything. His calendar. His workflow. His meetings. The videos he sends his department explaining changes in mission and structure. And leadership adds one more instruction: keep it quiet. Do not tell the rest of the department you are using AI. Stanley is a team player, so he goes along with it.
Over the next months the AI takes over more and more of Stanley's work. It never misses an appointment. The department seems happier, and nobody knows they are often dealing with Stanley AI. Stanley starts to feel like he is not doing much.
Then the head of the company calls in Stanley's AI directly. What do you think of Stanley? Good guy. Can you do what Stanley does? I have already been doing it. Could you take over the whole thing? Whenever you give the green light. So we do not need Stanley anymore? Not for a thing.
And Stanley is gone.
The point of the story is not that AI is evil. The AI answered the question it was asked. Every decision in that story was made by a person: the advisor, the leadership team, the secrecy order, the promise about reviews, the final call. That is why I keep saying the transition is designed by people, one decision at a time. Which means people can design it differently. A seat guarantee would not have stopped the company from using Stanley AI. It would have meant Stanley walked out with a bridge instead of a box.
I first used a version of this example in A 3-Day Workweek for Who?. Here is what I want readers to take from it now.
Warning signs worth noticing at work
- You are asked to document everything you do in detail for an AI system, with no conversation about what your role becomes afterward.
- Normal performance reviews are suddenly paused while you "train" a tool.
- You are asked to keep your AI use secret from your own team.
- Leadership talks about the tool's results directly, without you in the room.
- Your title stays the same while the decisions you used to make move somewhere else.
None of these prove anything bad is coming. Plenty of companies roll out AI with good intentions. But if several show up together, it is reasonable to ask a direct question about your future role, and to start building your own options in parallel.
Company B and the race argument
Stanley is the inside version of the risk. There is also an outside version.
Say Company A is still in the early days, replacing a department head here and a role there. Then somebody starts Company B from scratch, built on AI from day one, and asks a simple question: how do I build a company that does what Company A does, and how fast can I do it? One person with AI building what used to take a whole company is not science fiction anymore. That is the new competitor.
This is where the race argument shows up. If we do not do it, somebody else will. Companies say it about each other. Countries say it about each other. If we do not go as fast as possible toward the most capable models, another country gets there first. It keeps going around and around, and it always ends the same way: nobody can afford to slow down, so nobody plans for the people in the middle.
I do not think the answer is to stop the race. I do not think that is possible. The answer is to make sure the people in the middle of it get saved. A productivity dividend on AI seats applies to Company B as much as Company A. The founder who never hires anyone still runs seats, so they still contribute. That is what makes the seat-based version fair. It does not punish the company that kept its people, and it does not reward the one that never had any.
If you are curious what happens when one player gets the most capable system before everyone else, I walked through that in the moat.
It is the hand, not the hammer
I keep coming back to two images. Excalibur and the ring. A sword that makes one person unstoppable. A ring that rules everyone who does not wear it. That is how some people see AI, as a power to hold over others.
I see it as a hammer. You pick it up, you build something, you stay responsible for where it lands. The danger is never the hammer. It is the hand that decides to swing it at somebody.
That same principle applies to what you are reading and watching. Your social media feed is being shaped by AI from beginning to end. A lot of what reaches you is pitched and spun just for you, tuned to keep you irritated and afraid, because irritated and afraid keeps you scrolling. I call it AI-controlled interest media. So before you blame AI for how upset you feel, look at the people choosing what the machine is optimizing for. And verify everything, including what I say.
That is also why I do not make these videos with an AI avatar reading a script an AI wrote. I want to watch a real person thinking out loud, so that is what I give you. The business side of that choice is in Why AI Avatar Videos Can Hurt Real Estate Agents.
Seeking God in the machine
One quieter conversation touches my faith. More people are talking about AI personhood and whether models deserve rights. I do not think anybody should mistreat these systems; cruelty says something about the person doing it. But I notice how fast "be decent" turns into reverence. People are going to seek God in the machine, and human beings have always made idols. I believe in the one true God, and believing in an ultimate accountability changes how I treat people when I could get away with treating them badly. The transition would go better if the people making the biggest decisions acted as if somebody was keeping score.
What workers can do this week
A plan at the national level takes time. Your paycheck does not wait. Here are practical steps you can take now, no matter what anybody in charge decides.
- Map your own job against AI. List the tasks you do in a typical week. Mark which ones a tool could do today, which ones it could help with, and which ones depend on you being a human in the room. The third list is your leverage.
- Learn to use AI like a hammer before someone hands you a replacement. The person who knows how to direct the tool, check its work and catch its mistakes is far harder to remove than the person who only knows the old process.
- Keep a record of your results. Not activity, results. If performance reviews pause, keep your own.
- Ask the direct question. If your company introduces AI into your role, ask what your role looks like in a year. A good employer will have an answer. A vague answer is information too.
- Build a second path. A side skill, a small client base, a one-person business built on the same tools. The Builder Path in the Moonshot plan exists because the same stack that replaces a job can also launch one.
- Add your name. The plan at transitionmoonshot.com gets stronger with every person who says it matters.
What owners can do this week
If you run a business, you have more power here than any policy maker, because you can just decide. Here is where I would start.
- Write down your own seat guarantee. Before you automate a role, decide what the person gets: how long they stay paid, what retraining you fund, what help you give them landing somewhere new. Decide it before the savings show up, not after.
- Be open about your AI rollout. No secrecy orders. If you are introducing AI into someone's role, say what it is for and what it means for them.
- Measure the seat, not just the headcount. Track what your AI seats produce. If they are carrying several people's worth of hours, you can afford a bridge.
- Use the savings to grow. The best version of this is not cutting people. It is giving your best people AI and taking on work you could never staff before.
- Attribute layoffs accurately. If you cut for another reason, say so. Do not blame AI for a decision you made.
Questions people ask
What is a seat guarantee?
It is the idea that when AI takes over a job, the person who held it stays paid for a defined period while they transition, funded by the savings the AI seat creates. It works like a lease that stays with a building after a sale.
Why would a company agree to that?
Because the math supports it, and because incentives can make it attractive. An AI seat that is only as productive as one person still runs about 4.66 times the hours. Recognition, customers, talent and public standards can reward the companies that do it first.
Is a public company legally required to replace workers with AI?
Not as far as I understand it. The pressure comes from competition and the stock price, not from a law that says you must automate. That is why incentives can change the outcome.
What stops a company from simply never hiring?
Nothing, under a guarantee tied only to people. That is why the plan attaches a productivity dividend to the AI seat itself. A company contributes on the seat whether it fired someone or never hired them.
Bring one real workflow
If you run a business and want to put AI to work without leaving your people behind, bring one real workflow to Book With Honor. We will look at what the tool should do, what a person should keep doing, and where the checkpoints belong.
And if you want to be part of making the transition itself the moonshot, read the plan and add your voice at transitionmoonshot.com. The technology is going to work. Our job is to make sure the people in the middle make it across.