Brookings: No Silver Bullet for AI and Jobs — But Four Real Levers

About three-quarters of American adults say they fear AI will cost people their jobs. Fewer than 6% of private-sector workers belong to a union that could bargain over how that AI actually gets deployed. Hold those two numbers side by side and you get the most useful — and, oddly, the most hopeful — idea in a new Brookings framework published on June 29, 2026: the problem is not that nobody is worried. The problem is that the worry has had nowhere to go.

That is a fixable thing. And the paper is, in large part, a map of the people already fixing it.

One report, one argument

"Getting to all-of-the-above: A framework of solutions for AI's coming impacts on work and workers," by Xavier de Souza Briggs, senior fellow at Brookings Metro, makes a deceptively simple claim: there is no single policy that protects your job, and the search for one is precisely why so little has been done [Claim]. Briggs calls the framing of the debate — innovation or an inclusive future — "a false and dangerous binary choice."

So instead of picking a winner, he sorts every serious proposal on the table into four families and argues that a real response needs all of them at once. Hence the title.

The four levers

Brakes slow automation where the stakes are highest, without banning the technology. This is not theoretical. California's 2024 "human-in-the-loop" law requires human judgment in algorithmic decisions that touch sensitive outcomes [Fact]. Illinois went further in 2025, prohibiting AI from replacing community college faculty and licensed mental health professionals [Fact]. Several states have proposed automation impact assessments — a requirement that an employer study the workforce effects before the system goes live, not after the layoffs.

Steers are incentives: nudging employers toward pro-worker AI design, and nudging workers toward occupations with durable demand. Briggs is blunt that the US does this badly, underinvesting in workforce development relative to peer economies.

Buffers are the safety net, modernized: unemployment insurance built for a world where a job disappears rather than pauses, wage insurance for people who land in a lower-paying role, and training that is funded to an outcome rather than to a certificate. New Jersey has moved on "right-to-retraining" legislation; New York created a FutureWorks Commission to plan for worker economic security [Fact]. At the federal level, a bipartisan AI-Related Job Impacts Clarity Act would at least force disclosure — right now, nobody reliably counts AI-driven job losses separately from ordinary churn.

Shifts are the structural ones nobody wants to say out loud in a hearing: shorter workweeks, wealth-sharing mechanisms, guaranteed income experiments, and rebalancing a tax code that currently treats a machine more favorably than a worker.

Pace it. Redirect it. Catch the people it drops. Then change the shape of the deal itself. Four levers, four timeframes.

The "great mismatch" — the number that reframes everything

Here is the finding that deserves far more attention than it is getting. Brookings' own 2024 analysis of over 1,000 occupations found that a job's exposure to AI is inversely correlated with the odds that the person doing it belongs to a union [Fact]. Briggs calls it "the great mismatch": the workers most exposed are the least collectively organized.

Read that again, because it explains a lot of what feels irrational about this moment. The people with the strongest institutional voice at work — many skilled trades, public educators, nurses in organized systems — tend to sit in jobs AI touches least and latest. The people most exposed — administrative professionals, marketing and media workers, entry-level analysts, customer support — usually have no mechanism at all to negotiate over how a tool gets deployed on their desk.

That is not a doom statistic. It is a to-do list. It says the anxiety and the leverage are sitting in different rooms of the same building, and connecting them is a solvable engineering problem — not a law of physics.

Meanwhile the entry-level door is the one visibly narrowing. A 2024 analysis cited in the report found 55% of business leaders and 68% of investors expect less entry-level hiring [Fact]. Briggs' concern is specifically about "gateway" jobs — the roles that historically carried people without a four-year degree into higher-wage careers. Lose the bottom rung and the ladder still stands, but nobody new gets on it.

The corporate signal is loud, too: Meta announced 700 layoffs in March 2026 and roughly 8,000 more in May 2026, citing a pivot to artificial intelligence [Fact] — while the labs building that AI, OpenAI and Anthropic, are preparing IPOs at valuations reported around $1 trillion or more [Estimate]. The value is being created. The question the report keeps returning to is simply who is at the table when it gets divided.

The uncomfortable sentence: retraining alone will not save you

If you read one line from this report, make it this one: "Training or retraining alone does not produce good jobs, at least not directly and readily."

This cuts directly against the advice most workers have been handed for three years — take the course, get the certificate, you will be fine. Briggs, drawing on the hard lessons of deindustrialization, calls the American reflex "train-and-pray." Skills without demand, without employer commitment, without placement, without a wage floor at the other end, produce a credential and not a career.

This is not a reason to stop learning. It is a reason to stop treating learning as the whole plan. Training is necessary. It has never been sufficient. The countries that manage transitions well — Denmark's "flexicurity" model is the one Briggs points to — pair training with income support and active placement, so the worker is carried across the gap rather than handed a map and wished luck.

What this means for your job

The report's own read on exposure is uneven, and that unevenness is your best planning tool.

Most exposed, least protected: college-educated white-collar work — see our data on administrative assistants, customer service representatives, journalists, and paralegals, where task-level automation potential runs high and collective bargaining coverage runs near zero.

Chronic shortages, structurally durable: skilled nursing and the construction trades. Briggs argues these should be the target of coordinated public-private recruitment, not an afterthought — see registered nurses and electricians. And truck drivers, often used as the poster child of automation, appear in the report as an example of how uneven and slow the real timeline is.

The unifying point: your exposure is a fact about your tasks, but your outcome is a fact about the deployment decision — and deployment decisions are made by people who can be influenced. As Briggs puts it, "employers in every sector have a broad, not narrow, spectrum of choices about how to adopt AI."

What you can actually do this quarter

Find out what your state is doing. Illinois, California, New Jersey and New York are already legislating. State law is moving faster than federal law and it is far easier to influence.

Ask about the deployment, not the technology. "Will there be a human in the loop?" and "Has anyone assessed the workforce impact before rollout?" are questions with policy precedent behind them now. They are also, notably, questions an employer can answer without admitting anything.

Defend the bottom rung where you work. If your team is quietly not backfilling junior roles, that is the gateway-job problem happening in miniature — and it is usually a decision, not a fate.

Keep training, but demand a destination. Pair any reskilling with a real signal of demand: an employer commitment, a placement rate, a hiring pipeline. Train-and-pray is not a strategy.

The hopeful read

The report's timeline is not an apocalypse. Briggs expects disruption to unfold over decades, unevenly across regions, sectors and roles — not as one catastrophic morning. That matters, because decades are enough time for policy to catch up, and the catching-up has visibly started: state experiments, philanthropic initiatives like the Windfall Trust and Humanity AI, mayors convening on AI, and new measurement projects at Stanford and Yale's Budget Lab designed to tell us what is actually happening rather than what we fear.

The quiet good news in this paper is that the debate finally has structure. A year ago the choice on offer was panic or denial. Now there are four levers, a growing list of states pulling them, and a clear description of the gap — the great mismatch — that has to be closed. Fear needed a plan. This is the beginning of one.

Sources

• Brookings Metro — Getting to all-of-the-above: A framework of solutions for AI's coming impacts on work and workers (Xavier de Souza Briggs, June 29, 2026)
• US Bureau of Labor Statistics — Union Members Summary (private-sector union membership just under 6%)
• Stanford HAI — AI Index Report (cross-country AI sentiment)
• Anthropic — Anthropic Economic Index (task-level AI usage measurement)
• The Budget Lab at Yale — AI and the labor market tracking (measurement of AI's labor-market effects)

Related reading: Challenger June 2026: AI Leads Job Cuts for a Fourth Straight Month · ECB: AI Already Cut High-Risk Jobs by 4% While Safe Jobs Grew 13%

_This analysis was produced with AI assistance and reviewed by a human editor. Occupation-level exposure and automation figures come from our own dataset built on O*NET task structures._