Twenty-six weeks of wages for AI job loss. Then what?

On September 24, Senator Mark Kelly introduced the Make AI Work for Americans Act. It sets up a federal trust called the AI Horizon Fund, paid for by three new taxes on the biggest tech companies: one on excess profits, one on digital ad services, and one on AI usage. The headline benefit is simple. A worker who loses a job to AI could collect 75% of their old wages for 26 weeks. State unemployment usually replaces 30 to 50%. Microsoft and the Teamsters both back it, which is not a pairing you see every week.

In a TV interview about the bill, Kelly was candid that the job losses haven’t come as fast as people predicted. His case is that Congress is slow, so the fund should exist before it’s needed. That’s a sensible reason to act now.

A lot of this bill is right. It pays people, not programs. The replacement rate is high enough to cover rent. And it gets built before the crisis instead of during one.

The trouble is the number 26.

Unemployment insurance was built for layoffs

The whole design of unemployment insurance assumes a particular kind of job loss. A plant slows down, a store closes, a company has a bad year. You lose your job, and a similar job exists somewhere else, or will again when the economy picks up. The benefit is a bridge to that next job. Six months is usually long enough to cross it.

AI job loss in customer service does not look like that.

When a company moves most of its password resets, order status questions, and return requests to an AI agent, those seats don’t come back in the next upturn. The work didn’t pause. It moved. And the savings from moving it show up every quarter, for as long as the company runs the software. For the business, the savings are permanent. For the worker, the help ends at week 26.

AI has lately been the most common reason companies give for cuts. In its July report, Challenger, Gray & Christmas found that AI led all reasons for announced job cuts for the fifth month in a row, with 112,713 cuts attributed to it so far this year, about a quarter of the total.

To be fair about it, most people who lose a job do find another one. The wave of AI job losses will crest at some point. But a bridge only works if there’s land on the other side. When the whole category of entry-level support work is shrinking, the next job takes longer to find, and it often pays less. Six months of help doesn’t match a change that lasts for years.

Pair the bridge with a stake

The fix is not a longer check. Longer benefits get expensive, and they get cut the first time a budget is tight.

The fix is giving displaced workers a share of the thing that replaced them. The savings from automation don’t run out. A worker who owns a piece of them keeps getting paid after week 26, in the years when the new job pays less or takes longer to land.

That’s the idea behind The Dividend Standard. We sell customer service automation, so our product is exactly the kind that displaces support jobs. We commit the greater of 5% of our qualifying revenue or 70% of our adjusted profit, every year, to the workers our software displaces. It’s granted as ownership held in trust and paid out over years, not as a one-time payment. We’re a single company, and we only see a slice of what our customers save, so we can’t make anyone whole alone. That’s why the standard is written down in public, where other companies and lawmakers can copy it.

Kelly’s fund and a standing stake are not rivals. One covers the first six months. The other covers what comes after.

What a business owner can do before any of this passes

Congress is slow, as Kelly said himself. A business owner rolling out AI support this year doesn’t have to wait for it.

Displacement is fastest when a company flips a switch and moves everything at once. It’s slower, and easier to plan around, when the agent takes over one kind of question at a time. That’s how we built Celeste. It practices on your real support history first, then goes live topic by topic as you approve each one. You can watch which hours actually drop, and when.

That gives you time to move people into the work that is growing: the exceptions, the angry customers, the approvals the agent is not allowed to make on its own. Some roles will still go. But deciding that with six months of data is very different from deciding it in a board meeting.

Kelly’s bill would build a bridge. Somebody still has to build the other side, and the businesses keeping the savings are in the best position to do it.