The Great Un-Re-Bundling
I have spent the past year watching a set of macro changes that are usually discussed separately: the trade order re-routing along political lines, firms redrawing what they own and where they operate, AI absorbing tasks, careers going plural. They are interdependent, and from a workplace perspective the implications compound rather than add. Instead of continuing to turn this over privately, I decided to write it down and invite conversation, and disagreement. That is what this series is for.
For decades after World War II, a job was more than just a source of income. The employment relationship bundled pay, health insurance, pensions, training, identity, and community into a single relationship with one employer. Employer-provided health insurance spread because wartime wage controls barred raises, so firms competed on benefits instead, and a 1954 IRS ruling made premium contributions tax-exempt. Pensions and internal training programs evolved from the same expectation of long-term tenure. Identity and community became side-effects of years of working alongside the same people.
As Ronald Coase argued in 1937, firms exist because using the market carries costs of its own: discovering prices, negotiating contracts, enforcing them. Activities are performed by the firm when managing them internally is cheaper than contracting for it. For most of the postwar era, direct employment was the cheaper option, and every component of the bundle depended on it.
Digitization, offshoring, and platform intermediation changed the math. Once a task's output could be specified, measured, and enforced at a distance, the market could handle it, and firms outsourced the work. They externalized the risks as well: wage volatility, benefits, and liability moved onto workers and subcontractors. David Weil's The Fissured Workplace documents the result: firms shed direct employment to subcontractors, franchisees, staffing agencies, and platforms. The bundle's components were attached to the employment relationship for different reasons (tax policy, administrative scale, long tenure), but all of them depended on it. As direct employment declined, the benefits, training, and community it created around employment declined as well.
This is the Great Unbundling, and it is only half the story.
Unbundled components tend to reassemble around new anchors. Substack reassembled the audience and payment pieces that blogging had scattered. Apple has spent forty years rebundling hardware, software, services, and identity. The labor market is going through the same cycle. Who anchors the new bundle will determine what workers get from it.
Platform companies, employers, and community institutions are each trying to reassemble it, on different terms.
Platform companies are rebundling around themselves. Uber, DoorDash, and their peers offer work, ratings, insurance products, and somewhat of a community, all mediated by algorithms the worker has no ability to own or influence. Workers get a version of a bundle, but the platform keeps the surplus and sets the terms.
Some employers are rebundling to retain people. Walmart's Live Better U pays full tuition and books. Chipotle covers debt-free degrees. Target's Dream to Be and Amazon's Career Choice extend into graduate degrees and industry certifications. These are retention plays in a tight labor market, and they may not survive the next downturn.
Community and civic institutions are the third contender: workforce boards, community colleges, unions, and local nonprofits reassembling training, credentials, and job connections around a place rather than a single employer. They work with far less money than the platforms and large employers. Federal workforce formula funding under WIOA runs about $3 billion a year for the whole country, flat for a decade. Amazon alone committed $2.5 billion to its training fund.
These efforts differ in who keeps the surplus: platforms concentrate it in themselves, employer programs last only as long as the labor market stays tight, and only the community version is organized around the worker. The outcomes will differ by industry and region.
The workforce development field's usual answer is more public-private-civic collaboration. In every case where cross-sector collaboration has worked at speed, one actor paid the coordination cost up front: the legal work, the platform build, the political risk of going first.
Tulsa Remote, launched in 2018 by the George Kaiser Family Foundation, paid remote workers to relocate. The Economic Innovation Group estimated the program generated roughly $13 in new local labor income for every incentive dollar spent. The foundation took on the upfront cost and the reputational risk of paying people to move, a bet no city council was likely to approve.
People+Work Connect, launched in April 2020 by CHROs from Accenture, Verizon, ServiceNow, and Lincoln Financial, went from idea to platform in fourteen business days. It moved workers between companies as some firms laid off retail staff and others hired warehouse workers as fast as they could onboard them. Accenture built the platform at its own cost, and a small group of CHROs vouched for it with their peers. Companies invited afterward took on almost none of the setup cost.
RAISE US, launched in 2026 and co-led by former Commerce Secretary Gina Raimondo and former Indiana Governor Eric Holcomb, applies the same model at national scale for the AI transition. It has raised roughly $500 million toward a $1 billion target from anchor partners including OpenAI, Anthropic, Microsoft, and Amazon, plus employers (Cisco, IBM, Cognizant, Mastercard, Bank of America), governors (Maryland, Arkansas, Utah, Connecticut), labor (AFL-CIO), education (WGU, ASU, Khan Academy, Jobs for the Future), and research (Burning Glass Institute, MIT's David Autor, Harvard's Raj Chetty). Raimondo and Holcomb take on the political work of getting business and labor to the same table, and the AI companies provide most of the money.
RAISE US is also funded largely by the same frontier AI labs and large employers whose products are displacing the workers the coalition aims to serve. They have the money and also the responsibility to respond, but a coalition funded by the disruptors leaves open the question of who the rebundling ultimately serves. The funders' interests and the workers' interests won't automatically align and will require intentional design. David Autor, on the RAISE US board, has documented how the China Shock devastated manufacturing communities while federal retraining programs reached too few workers, too late. RAISE US is a bet that the response can be built faster this time. Whether it delivers for workers will take years to find out.
National coalitions can build standards, funding, and political cover, but they cannot reach individual workers on their own. That requires place-based ecosystems: workforce boards, community colleges, apprenticeship intermediaries, and civic conveners like CIV:LAB, the local institutions that know the people and the employers involved. Federal programs have lost momentum at that handoff for decades.
AI speeds up both the unbundling and the rebundling. It makes work easier to break into tasks and route to whoever does them cheapest. It also lowers the cost of the coordination systems anchors need, matching workers to openings, benefits to eligibility, training to demand signals, work that used to require large institutions and heavy investment. Who owns that infrastructure will shape whether AI mostly helps the platforms or mostly helps the workers.
The rebundling is already underway. Who anchors it, on whose behalf, and how fast: I'll take those questions up in turn in the articles that follow. This is the first piece of a longer series that looks at the same un-re-bundling through different lenses: the firm as the trade map changes, work itself as AI absorbs tasks, the job description, the HR systems built on top of the job, and the career as the bundle the individual ends up carrying.
If you see these changes differently, or see ones I am missing, say so. I would love for this to be a conversation so please opine.
