Our commitment is fixed. The standard is an invitation. A company that profits from automating people's jobs owes something real to the people whose jobs it automated. Our share is the greater of 5% of our qualifying revenue or 70% of our adjusted profit, every year, granted to displaced workers as ownership. That is settled and it is ours to keep whether or not anyone joins us. What we are inviting others to build is the rest: the industry standard, the safe harbor in law, and the coalition of companies that would let all of us give back far more than one company can alone.
Why we're doing this
We build software that does work people are paid to do. We are not going to pretend otherwise, and we are not going to pretend it lands softly on the people it replaces.
AI is going to create an enormous amount of wealth. The question that matters is who ends up holding it. If nothing changes, the answer is the companies that build and buy these systems, along with the investors behind them. The person whose job went away gets a severance check, if they are lucky, and a note wishing them well.
We build these systems and we make money doing it, so this is ours to help fix. We commit the greater of 5% of our qualifying revenue or 70% of our adjusted profit, every year, to the people our products put out of work, granted as ownership. This page shows how that works and asks the rest of the industry to adopt the same standard.
One hard fact sits underneath all of it. No single company can solve this alone, and the reason is arithmetic, not nerve. When our software replaces a job, most of the money that job used to cost never reaches us. It stays with the business that bought the software and stopped paying the salary. We only ever see a slice of it. So our profit on its own, even every cent of it, cannot make a displaced worker whole. That is why we are proposing a standard for the whole industry rather than quietly writing our own checks.
What we believe
A few beliefs sit under the plan.
We'll publish what reached people. Claims filed, claims approved, and workers paid, in aggregate, year after year. And as the fund grows, an independent accountant will confirm annually that the formula was paid in full, so the arithmetic never rests on our word.
The workers' share should be ownership, not a handout. A check can be stopped the first year money gets tight. A stake in the company is much harder to take back, and it keeps paying as the business grows. So the goal is to give displaced workers a genuine share, held somewhere a future board cannot quietly undo.
The need is temporary, but the money it draws on is not. The wave of lost jobs will eventually crest. The profits of an automated economy will not. That gap is the reason any of this can work, because a fund fed by lasting profits can keep paying people long after the hardest part of the transition has passed.
It only holds if others join. On our own, we can return very little before a competitor who returns nothing undercuts us. Every company that adopts the same standard raises the amount all of us can afford to give back. It is one of the few commitments that gets easier to keep the more people make it.
What we're proposing
Here is how our own commitment works. The number is set: every year we direct the greater of two figures to the people our products displace.
The greater of 5% of revenue or 70% of profit. Five percent of our qualifying revenue, or seventy percent of our adjusted profit, whichever is larger that year. The revenue floor is the part that matters most in the early years, because a company can reinvest everything and show almost no profit; a pledge tied only to profit can quietly pay nothing. Tied to revenue, real money flows even in a year we lose money. As we mature and margins grow, the profit figure overtakes the floor and the commitment rises with it. Both figures are defined against our audited financials, with mandatory add-backs so profit cannot be suppressed by paying insiders above market or dressing up owner distributions as costs. The one exception: in a year we could not pay the floor without threatening the company's survival, the shortfall carries to the next profitable year and we say so publicly. That is the only deferral, and it never applies to the profit figure.
The commitment is in force now, and the structure around it is honestly unfinished. We are not waiting for a perfect trust and a polished claims process before the first payment. If our software cost you your job, file a claim today. The rules we apply are published below.
A trust that holds the promise. We are placing a special class of shares into an irrevocable trust whose only job is this. Those shares carry the right to the pledge we pay out and a share of the proceeds if the company is ever sold, codified in our charter as a public benefit corporation and changeable only by a supermajority of both the board and every class of shareholders. The point is to turn the commitment into a property right that outlasts us and cannot be voted away.
A stake for each displaced worker. Someone whose job our software replaces would receive units in that trust, sized to the pay they lost, and paid out over years rather than all at once. Two honest things about what a unit is worth. The yearly cash comes from the pledge pool, so it does not depend on our shares ever being worth anything. The ownership half does: those units ride on the fortunes of the very company that displaced you, and if we fail, that part fails with us. We would rather name that plainly here than have you find it in the fine print.
An invitation to the businesses that save. Most of the money automation frees up stays with the company that bought the software. So we ask our customers to consider passing a share of their verified savings into the fund, and we provide the pathway when they want it. We cannot require this, and we do not. Buying Celeste obligates you to your invoice and nothing else: the pledge above is funded from our share of what you already pay, with no surcharge, no line item, and no bill that arrives later. The contribution is a choice, it would start small and rise as more of the industry takes part, and the buyer who makes it still comes out well ahead of the salary they stopped paying. The worker gets a ramp instead of a cliff.
One number to judge us by. The standard comes down to a single figure: for every dollar of wages our software displaces, how many cents actually reach the worker who lost them. A company acting alone can only fund a small part of that dollar, because most of the displaced salary stays with the business that bought the software, not with us. On our own, our commitment funds on the order of ten to fifteen cents of every displaced dollar. With buyers contributing under a shared standard, that reaches toward sixty. The gap between those two numbers is the entire reason we are asking the rest of the industry to join. A company doing none of this returns nothing at all.
How we decide who qualifies
We cannot prove that any single job was lost to our software and nothing else. Work rarely comes apart that cleanly. What we can do is set a clear standard, apply it the same way every time, and err toward including people. We do not have all of this worked out, and we would rather run an honest, imperfect version now than a polished one someday. These are the rules we hold ourselves to today, and they will get more rigorous as the fund grows.
Three things have to be true. First, you are a real, unique person: every claim starts with identity verification, confirming you are who you say you are and that one person is filing once. Passing that step only lets the claim move forward. Second, you worked in a covered role at a company that ran Celeste. A covered role is front-line service work of the kind our software performs: customer support, contact center, help desk, and equivalent roles. A pay stub, W-2, offer letter, or employment records shared with your consent shows this, and we check the employer against our own records of where Celeste ran and when it went live. Third, you had a qualifying separation inside the window, which runs from the day Celeste went live at your employer to twelve months after. A qualifying separation is a layoff, an eliminated position, or a role reduced out from under you. A government unemployment determination is the strongest single proof, because a public agency has already reviewed it; a separation letter or severance agreement also works.
When all three line up, eligibility is presumed. You are in unless there is a specific, documented reason you are not.
Your former employer can rebut, never approve. We do not make you get the blessing of the company that automated your job; that would put your standing in the hands of the party least likely to help. We notify them and give them a set window to object with cause: you left voluntarily for a better job, you were separated for documented misconduct that predates the deployment, or your role had nothing to do with what Celeste does. Silence is not a rejection. If they say nothing, the presumption stands and you are in.
What stops fraud. A false claim does not steal from us. It steals from a displaced worker, because every unit that goes to the wrong person never reaches the right one. So identity checks catch duplicates, employment and separation evidence is checked against records rather than taken on someone's word, we audit a sample of approved claims in full, and a claim found fraudulent is denied, clawed back, and, when the fraud is clear, referred for prosecution.
Where this is a proxy, we say so. A covered role, a Celeste deployment, and a separation inside the window is our best honest stand-in for a cause nobody can measure directly. It will sometimes include a person whose exit had another reason, and it will sometimes miss a person whose loss was real but whose paperwork does not fit. We would rather the error fall toward including people. We also know the deeper conflict: today, the company paying the claims is the company reviewing them. As the fund grows we intend to put those decisions in independent hands. Until then the rules are public, the outcomes will be published in aggregate, and when the standard gets something wrong we would rather fix it than defend it.
To the worker whose job we automated
You did nothing wrong. The work changed underneath you, the way it changed for the people who ran switchboards and set type before you. The one difference we are trying to make is that this time the people who built the machine owe you something real and put it in writing.
We are not going to tell you this makes you whole. In the best honest case, what these pieces add up to is several years of your pay, spread across the years ahead. That is real money that can change a life, and it is still less than the job itself would have paid you over the same time. What it is meant to be is a funded transition and a lasting stake in the economy that replaced your job. What it improves on is what workers have been handed every time before this, which was a severance check and good luck.
To the CEO deciding whether to join
We are not going to appeal to your conscience. The case is in the numbers. Every company that automates away workers is, across the economy as a whole, automating away its own customers. A person who has lost their income and has no clear way to a new one cannot buy what you sell. A standard like this keeps money in the hands of the people you are counting on to be your market.
Joining also changes what you can afford to give. On your own, you can return only a little before a rival who returns nothing undercuts you on price. When many companies adopt the same standard, that pressure eases and the amount each of you can sustain goes up. A smaller share of a company whose customers can still pay is worth more than the whole of one whose customers cannot.
To the lawmaker who can change the rules
This is one of the rare problems where fixing it costs less than leaving it alone. A few pieces need you specifically. Securities law already has clean paths for a company to grant shares to its own employees. It has no clean path for what this plan needs, which is granting them to our customers' former workers, people who never worked for us at all. That gap is why the rules would need a narrow and careful exemption.
The harder piece is antitrust. The laws that make price fixing illegal, the ones that stop competitors from quietly agreeing on what they charge, would also catch a group of AI companies agreeing to add the same transition contribution. It does not matter that the aim is to fund displaced workers rather than to raise profits. On paper, rivals coordinating on a shared charge looks close enough to a cartel that the risk of a lawsuit alone would keep any careful company from joining. A law written to protect customers would end up blocking one of the few things that could actually help them.
The fix is a narrow safe harbor. Lawmakers have made room like this before, when coordination serves a clear public purpose, and the same tool fits here. A company that follows the rules of a certified transition standard would be shielded from antitrust liability for that one kind of coordination and nothing else. Because the protection is tied to the certification, it covers funding workers and never turns into a license to fix prices on what a company sells. That is also why certified has to mean something real, with a recognized body and rules a company actually has to meet.
You can also help carry the cost, because every worker moved along a funded transition is one who does not file for unemployment. And once the model has proven itself, you could set a floor: any employer who automates owes a contribution toward the transition, which they can satisfy by joining a certified standard. Until it is proven, we are asking for room to try, not a mandate.
The ask
Our own commitment is set and in force. What is not yet built is the thing bigger than us: an industry standard, a coalition, and the change in law that lets it scale. If you are a worker, hold us to our number. If you run a company, take the model apart, and when it holds up, join us. If you write law, give this room to work. If you study this for a living, tell us where it breaks.
Help us build the standard.
An open invitation from IMCeleste. We are one company starting this, not a coalition pretending to be one. It becomes a real standard only if other people help write it and agree to it.