Who keeps the time AI saves?

Quick Take · News · Put AI to Work · Looking Ahead

In this issue

AI is already giving knowledge workers hours back every week. The fight that actually matters is who keeps them: the company as bigger quotas, or the worker as a shorter week. Even OpenAI is now arguing for the four-day version.
Pope Leo XIV put the Vatican on record: "technology is never neutral." Religious institutions are turning into real coalition partners on AI and human dignity.
Three stories that show what we're up against, and the progressive move in each: AI-faked Black influencers, a startup filming domestic workers to train robots, and coders quietly de-skilling on AI.
A union win you can point to: Sports Illustrated's journalists just bargained real AI guardrails into their contract.
A Put AI to Work you can actually use this week: design a whole four-day-week pilot for your org, with AI as your planning partner.

Quick Take

Here's a number that should reframe the whole AI-and-jobs argument: AI already saves the average knowledge worker something like seven and a half hours a week. That time is real, and it's showing up now. The only open question is who keeps it.

There are two answers. The company keeps it, turns it into bigger targets, and the work just expands to fill the hours AI freed up. Or the worker keeps it, and a five-day job becomes a four-day one at the same pay. Nothing about the technology picks between those. People do.

The hopeful part: that argument is no longer coming only from labor. This spring OpenAI itself floated a four-day week as one way to share the gains. When the company building the technology and the workers who'd benefit are pointing the same direction, the progressive job is to make "shared with workers" the default instead of the exception.

Let's get into it.

92%

The share of companies that kept the four-day week after the UK's big pilot ended: 56 of the 61 that took part, with 18 making it permanent. Over the trial, run by the Autonomy Institute across roughly 2,900 workers, revenue held steady (up 1.4% on average), staff quitting dropped by 57%, and 71% of employees reported less burnout. Same output, less burnout, and nine in ten employers kept it going once they saw the numbers. The four-day week stopped being a thought experiment a while ago. AI is what makes the math work for far more workplaces now.

AI is handing us hours back. The fight is over who keeps them.

For two years the dominant AI-and-work story has been displacement: the jobs AI takes. There's a quieter story underneath it that's just as real: the time AI gives back. Tasks that used to eat an afternoon (a first draft, a data cleanup, a batch of code, a research summary) now take minutes. A 2025 study from the London School of Economics' Inclusion Initiative and the consulting firm Protiviti put it at around seven and a half hours a week for the average worker, close to a full workday.

That time doesn't disappear. It goes somewhere. Right now, in most workplaces, it goes straight back to the employer as higher expectations: same week, more output, and the productivity gain shows up as profit, not as time off. The alternative is to bank some of it: keep pay flat, drop a day, and let AI productivity fund a four-day week instead of a bigger quota.

The interesting development this spring is who's making that second argument. In April, OpenAI published a policy document, "Industrial Policy for the Intelligence Age," that proposes piloting a 32-hour, four-day workweek paid as a full 40, alongside ideas like profit-sharing and a public wealth fund. Sam Altman framed it as "a starting point, not a prescription." Take it with the appropriate grain of salt about the company's motives. But when the firm at the center of the AI boom is on record saying the gains should buy workers time, the old "pro-worker versus pro-AI" framing falls apart.

Why this matters

This is the most hopeful framing of AI and labor available right now, and it's one progressives are unusually well positioned to carry. The shorter-week movement already has the receipts: the pilots work, companies keep them, people are healthier and don't quit. What it didn't have was a forcing function big enough to make it mainstream. AI is that forcing function. Every hour AI saves is an hour that's now on the table to be claimed, by one side or the other.

And it reframes the entire conversation away from fear. "AI is coming for your job" is a defensive crouch. "AI should be giving you your Fridays back" is something to organize toward. Same technology, completely different politics. The difference is whether workers have a claim on the gains, and whether anyone's making that claim out loud.

This is the work Four Days With AI exists to do. It's a small site I put together making exactly this case, with a practical roadmap for moving an organization from "AI saves us time" to "AI bought us a day." More on that in Put AI to Work below.

What you can do

Start naming the gains out loud. When AI saves your team time, the default is silence: the hours quietly get absorbed and nobody decides anything. Make it a decision instead. In your next team or board conversation about an AI tool, ask the question directly: "If this saves us X hours a week, where do those hours go?" Putting it on the table is most of the battle, because the absorb-it-into-bigger-quotas outcome only wins when nobody asks.

The Pope put the Vatican on record about AI

Pope Leo XIV released an encyclical in May, Magnifica Humanitas, with a line worth holding onto: "technology is never neutral." It frames the AI moment as a choice between unchecked growth that leaves people behind and a path that, in his words, "rebuilds relationships before rebuilding with stones." He's explicit that when AI touches public goods and fundamental rights, it has to be governed by clear criteria and real oversight, and he calls for "courage and solidarity" in steering it toward people rather than concentrated power.

You don't have to be Catholic for this to matter. Religious institutions reach enormous numbers of people and carry moral authority that tech executives simply don't. On human dignity, worker rights, and accountability, that's a coalition partner progressives should be actively building with. The values line up more than the usual political map suggests.

What we're up against, and the move

Three stories from the week that show the other timeline, each with the progressive response attached.

AI-generated "Black influencers" are running scams. Reporting from The Verge documents scammers using AI to spin up fake Black women on TikTok (complete with invented hardship stories and fake appeals to racial solidarity) to sell cheap dropshipped products. It's digital identity theft at scale, and it undercuts real Black creators trying to build something authentic. The move: this is the strongest case going for content provenance and platform accountability, clear standards for labeling AI-generated personas, and real consequences for platforms that profit from letting them run. Progressives should be the ones writing those standards.

A startup wants to film domestic workers to train robots. Also via The Verge: a startup called Shift is offering free house cleaning in exchange for filming the work, footage that trains the household robots meant to do those jobs later. Workers (disproportionately women, immigrants, and people of color) become the unpaid training data for the systems built to replace them. The move: consent and compensation for training data. We saw the upside version of this in Issue 16, when Spotify and Universal built an AI-covers deal on "consent, credit, and compensation." The principle travels. If your labor trains the model, you have a claim on it, and domestic-worker organizations are exactly the groups that can press it.

Coders are starting to refuse work without AI. TechCrunch reports a growing dependence on AI coding tools, with researchers warning it can quietly erode the underlying skill and produce worse code over time. The move: the problem isn't the tool, it's the terms. The answer to de-skilling is bargaining for training and skill development alongside the AI, so workers come out of it more capable rather than more replaceable. Which is a good segue to this week's win.

Progressive AI Win

Sports Illustrated's journalists bargained AI guardrails, and won.

Here's what "bargaining for the terms" actually looks like when it works. In May, the 64 journalists at Sports Illustrated, represented by the NewsGuild of New York, ratified a new three-year contract with publisher Minute Media after roughly a year and a half at the table. The headline wins are solid: a $70,000 salary floor, an average raise of 5.22% on ratification with 3% guaranteed in each of the next two years, and preserved just-cause protections.

The part worth celebrating here is the AI language. The contract puts real guardrails on how the company can use AI in the newsroom: exactly the kind of "you don't get to spring this on us, and you don't get to use it to quietly replace us" protection that doesn't exist unless workers negotiate it in. No federal law required, no waiting for a regulator. A union sat across from a publisher and wrote the rules itself.

That's the answer to the de-skilling worry up above, made concrete. AI guardrails aren't something you hope a company adopts. They're something you bargain for, and a mid-size newsroom just showed the rest of the industry it's winnable.

Put AI to Work

Practical ways progressives can use AI this week

Design a four-day-week pilot for your org, with AI as your planning partner

Past issues have walked through using AI to make something: a benefits guide, a fact-check, an audit tool. This week is different. The task is to use AI to design a plan: a real, presentable four-day-week pilot proposal for your organization. AI is good at exactly this kind of structured thinking, and the whole thing fits in an afternoon.

Load the real context first. Open Claude or ChatGPT and tell it the truth about your org: how many people, what they actually do, where the time goes, what leadership cares about (cost, retention, output, mission). The more specific you are, the less generic the plan. If it helps, point it at the roadmap on Four Days With AI (the time audit, banking half the saved hours, freezing quotas during the trial, a six-month review) and ask it to adapt that structure to your situation.

Build the time audit. Ask: "Design a two-week exercise to measure how much time AI tools are actually saving each role on my team, and where." You want a lightweight way to capture the gain, because the pilot's entire credibility rests on it. AI freed the hours; now you're proving it on paper.

Draft the proposal, and make it argue both sides. Have it write the pilot proposal: scope, which day, how pay and coverage work, what you measure, the review checkpoint. Then turn it adversarial: "Now make the strongest case against this pilot, the way a skeptical CFO would." Answer those objections inside the proposal. A plan that's already survived its own best counterargument is far harder to wave off.

Make the leadership pitch. Ask it to compress the proposal into a short deck or a one-page brief aimed at whoever signs off, leading with their priorities, not yours. (The site has a Gamma-based pitch template you can start from.)

Then a human owns it. AI drafted the plan; it doesn't get to approve it or run it. You bring it to your people, pressure-test the numbers against your real operation, and decide. The point of the exercise isn't a perfect document: it's walking into that conversation with a credible plan instead of a vague wish.

From our friends

Your org deserves its own AI. Not Big Tech's.

Change Agent is a private AI platform built for nonprofits, unions, and advocacy orgs. Your data stays yours, it plugs into tools you already use (Google Drive, Slack, ActBlue), and it handles the tedious stuff so your team can focus on the mission. Starts at $35/month. Small nonprofits under $1M can apply for discounted pricing.

Learn more

Looking Ahead

A few threads to watch.

The four-day-week conversation is about to get more concrete. With OpenAI putting it in a policy document and the pilot data piling up, watch for which states and employers actually run trials with the AI-productivity framing attached. That's where this moves from idea to precedent. If your own org is the one to try it, the Put AI to Work above is your starting point.

California's WARN Act recommendations, the 180-day clock that started May 21, are still ticking toward a report due around November. That's the test of whether last issue's workforce order has teeth.

And Apple's WWDC lands June 8. After two years as the conspicuous no-show in consumer AI, Apple putting generative features in front of a billion devices would reset what "everyday AI" means for a lot of people. We'll cover what's worth your attention and what's marketing.