Corporate training is being rebuilt from the ground up
For the past two years, the question people asked most about AI and work was: will AI take my job? By 2026, that question is giving way to a more practical one — can a company's own training system still be used as-is?
In April 2026, the World Economic Forum (WEF) shifted its focus from "how big is AI's impact" to "how to build AI-era skills at scale." Around the same time, the Brookings Institution stressed that the future of work can't fixate on technology alone and has to come back to people. None of this is new in itself, but the signal is clear: the institutional conversation has moved from individual anxiety to organizational action.
The problem is, there is no one-size-fits-all template for that action. Blue-collar and white-collar workers are not facing the same thing. Blue-collar workers hit a wall where "the job is still there, but the person lacks new skills." White-collar workers hit the opposite wall: "the person still has skills, but the career ladder beneath those skills is collapsing." Stuffing both groups into the same "reskilling" program is a way to fail both. What this article wants to unpack is how companies should split their training systems apart and rebuild them for each group.
One word, two paths
Start with blue-collar. A visible shift around 2026 is that experienced workers began seeking training on their own initiative. The Guardian reported on a group that fits this pattern: skilled, seasoned workers, some with a decade or two of experience, learning AI-related skills with an urgency that borders on anxiety, afraid of being left behind. This breaks the old assumption that skill anxiety is a young person's problem — the ones most unsettled are often people who suddenly realize their hard-won craft may get repriced.
White-collar is a different picture. In May 2026, the consulting industry sent a string of signals: McKinsey launched an "AI-era compensation reform" that cut partners' cash dividends; the Financial Times reported BCG, Bain, McKinsey, and the Big Four accounting firms collectively freezing starting salaries. 36Kr followed up on McKinsey's moves. This isn't a short-term, one-or-two-round layoff wobble — it's an industry redistributing the value inside its pyramid.
Put the two side by side and the mismatch becomes clear: blue-collar workers face a case where the requirement of the role has changed while the person still waits in the old job; white-collar workers face a case where the skills remain but the ranks and billing models that carry those skills are coming apart. The former lacks new capability; the latter lacks a new career structure. Answering both with the same "reskilling" program is like trying to open two different locks with one key.
This mismatch also explains why so much corporate training money goes nowhere. Give blue-collar workers a generic "AI 101" course that has nothing to do with the specific equipment and process on their floor, and they still can't use it when class ends. Give white-collar workers a course on operating a new tool, but don't touch the promotion and billing rules behind it, and they still can't climb. Training isn't impossible — it just has to start by knowing which kind of problem you're facing.
Blue-collar: training should follow the job, not the certificate
Whether blue-collar reskilling works depends less on how good the course is, and more on whether there is a still-living job waiting behind it.
The failure of many past reskilling programs was this: people finished a course, got a certificate, went back to the labor market, and found the matching roles had either disappeared or simply didn't ask for that piece of paper. A certificate isn't wrong in itself — what's wrong is that it's detached from the actual job. What actually works is to embed training inside the job: first figure out which roles still exist, which are changing, and which are newly emerging, then design skills along that line.
A trend that's taking shape is that companies are no longer running "one-off training," but building "AI learning paths" for employees: at different career stages, they attach matching skill modules so that learning evolves alongside the role. This isn't just "a few more classes" — it's treating skills as part of the job that has to be maintained.
Vocational education is also repositioning itself. Palantir's CEO made a controversial claim — that AI will destroy many humanities-type jobs, but "people with vocational training will have plenty of job opportunities" (Fortune). The statement is extreme and debatable, but one thing it points to is fair: in the AI era, skills that sit close to a concrete job and can be put to work immediately may command more leverage than a broad degree.
Policy is also probing. "Labor SHIFT," proposed by US Democratic gubernatorial candidate Francesca Hong, contains a set of worker-protection ideas for the AI era (Urban Milwaukee, 2026.04). It's still a state-level campaign proposal and can't be treated as settled policy, but it represents one policy-level response to blue-collar job protection.
The blue-collar path comes down to one line: the value of training is ultimately decided by the job exit behind it. If the job is there, training has value; if the job is gone, certificates pile up in vain.
White-collar: when reskilling itself stops working
The white-collar side is trickier, because the reskilling remedy that everyone pinned their hopes on is increasingly being judged by institutions as not enough.
Consulting is the best sample to observe this. Its business model has long rested on a pyramid: a large number of low-cost junior consultants do the analysis, run the numbers, and draw the slides, while a small number of senior partners allocate the rewards at the top. AI is exactly what undermines the base first. A joint MIT/CNBC study found AI can already substitute for roughly 11.7% of the US workforce, with an even higher rate in professional services like consulting, accounting, and law.
Executive predictions go further. The head of Microsoft AI (via Fortune) offered a timeline suggesting white-collar work could be automated at large scale within a relatively short window; Anthropic's CEO made a similar forecast. Axios's "Behind the Curtain" column went as far as the phrase "white-collar bloodbath." These are judgments, not settled facts, but together they reflect one thing: white-collar anxiety about AI is moving from "it might happen in the future" to "it's happening now."
McKinsey's own moves work as a footnote. Business Insider reported its CEO saying AI is reshaping the talent structure — some roles growing about 25%, others shrinking about 25%; CFO.com reported it is piloting AI-driven hiring that pressures the traditional "up-or-out" model. eFinancialCareers described this stretch as "the worst time to become a consultant." Sequoia Capital threw out an even bigger proposition: "Services: The New Software" — professional services themselves are becoming software, and the per-person-day pricing model loses its footing (WSJ made a similar observation).
So the real dilemma surfaces: if AI can replace entry-level analysis and mid-level judgment, what exactly is "reskilling" supposed to teach? Short courses can hardly teach what AI can't replace, while the truly scarce abilities — deep judgment, complex negotiation, strategic trade-offs — can't be built in a few months.
Yale Insights makes a sharper observation: much of AI's destruction of white-collar careers happens before those careers even begin. Graduates find entry-level roles disappearing, with the "learning by doing" that once trained them taken over directly by AI. That means the problem isn't only "should current workers reskill" — it's "which route do new people come in through."
The US National Academies of Sciences, Engineering, and Medicine confirmed this dilemma with government-grade rigor in "Training Workers for an AI-Enabled World"; Harvard Gazette posed it directly with "AI took your job — can reskilling help?" Several outlets — Fast Company, Inc., Observer — each ran independent pieces under almost identical headlines: "Retraining won't save us."
The white-collar path is structural: it isn't "learn one more skill," but rather rebuilding three things at once — the entry channel, the promotion ladder, and the billing model.
These three are interlocking. The entry channel is blocked because junior roles get taken over by AI and newcomers no longer have the low-level work to learn on. The promotion ladder collapses because the "up-or-out" mechanism, which relies on churn to fuel promotion, actually pushes people out faster once total headcount shrinks. The billing model loosens because clients become less willing to pay by the "person-day" when AI can handle analysis and modeling. Fix one in isolation and the other two pull the result back to where it started.
Four levers for rebuilding a corporate training system
Once the blue-collar / white-collar difference is laid out, a company's actions can be distilled into four things. For each one, blue-collar and white-collar demand different approaches that can't be blended.
First, course design: from "add one more class" to "redesign around the task." Blue-collar courses should be designed around concrete processes and job tasks, so people can go straight to work afterward. For white-collar, the issue isn't a single course — what needs rebuilding are capabilities like judgment and collaboration that resist fast acquisition. The first principle of course design is to admit up front that some abilities can't be taught, then decide which ones are worth teaching in a structured way.
Second, the skill map: draw "what the job needs" and "what the person needs" separately. A blue-collar skill map can spread along equipment, process, and safety, focusing on what transfers from an old role to a new one; a white-collar map has to answer the harder question of which layer a person should move toward once AI takes over certain basic work. The value of a skill map is turning the vague "learn something new" into a chart a person can locate themselves on.
Third, the transition path: give people a route they can actually walk, not just a certificate. A blue-collar transition path ends at a job that still exists; a white-collar path may end at a role that has been redefined. The hard part of path design is the middle steps — between the old role and the new one, what to learn first, what work to pick up first, and who walks alongside. Training without a path tends to become "certificate-issuing" busywork.
Fourth, organizational support: training is not one department's job. Blue-collar training has to connect with production, safety, and unions; white-collar training pulls in pay structure, promotion mechanisms, and hiring standards. If the organization still runs on old rules like per-person-day billing or up-or-out, no course design can land. The essence of organizational support is making the surrounding rules change together with the training.
Of these four, the first two lean toward "design," the last two toward "grounding." Where many companies get stuck isn't course design — it's that the rules outside the course didn't move with it.
One more thing that's easy to overlook: the pacing of these four levers can't be the same for blue-collar and white-collar. Blue-collar training can move in step with equipment upgrades and line changes, because there are concrete machines and processes to anchor on, so results show more directly. White-collar restructuring involves relationships between people and organizations — it's slower and harder, and usually needs a small pilot to validate before scaling. Expecting one injection of money to pay off within a quarter is unrealistic for either group, and especially for white-collar.
The endpoint of training isn't "using AI" — it's "working with AI"
Back to a more basic question: what is AI literacy, really.
If it's just "knowing how to use some AI tool," that's a skill-level thing, a few months' worth. The hard part comes after — whether a person can judge if what AI produced is right, knows when to trust it and when to stop, and can find their own place inside a "human plus AI" workflow. That's the core of AI literacy: not which buttons to press, but knowing how to get the job done well together with AI.
On this point, the endpoints for blue-collar and white-collar actually converge. Whether the work is physical or cognitive, the human-AI relationship is shifting from "replacement" to "recombination." The only difference is which hurdle each group has to clear: blue-collar workers have to get past "does the job still exist," white-collar workers have to get past "does the ladder still connect." Whoever sees their own hurdle clearly first is the one whose training moves from "issuing certificates" to "building a real passage."
The endpoint of a training system isn't teaching someone to operate a tool. It should leave a person, after AI has taken over more and more of the basic steps, still knowing clearly: where I stand, and where I can still go.
Read the other way, this sentence also speaks to companies. A training system is only reliable if it answers two questions: where am I steering people, and can I actually connect them to new roles and new ladders. However many courses you run and however good the instructors, if the answer to those two questions is empty, training is a cost, not an investment. AI hasn't canceled training — it has merely put the question of "what is training actually for" back in front of every company.
References
- World Economic Forum (WEF) — "Invest in the workforce for the AI age", 2026-04
- Brookings Institution — "A people-first vision for the future of work in the age of AI", 2026-04
- The Guardian — "Skilled older workers turn to AI training to stay afloat", 2026-04
- Fortune — Palantir CEO on AI and vocational training, 2026-04
- Urban Milwaukee — "Labor SHIFT Policy to Protect Workers in the AI Age", 2026-04
- Financial Times / 36Kr — McKinsey compensation reform and consulting salary freezes, 2026-05
- National Academies of Sciences, Engineering, and Medicine — "Training Workers for an AI-Enabled World", 2026
- Harvard Gazette — "AI took your job — can retraining help?", 2026
- Yale Insights — AI employment destruction happens before careers begin, 2026
- MIT / CNBC joint study — AI substituting share of the US workforce, 2026
- Fortune — Microsoft AI head on white-collar automation timeline, 2026
- Sequoia Capital — "Services: The New Software", 2026
- Axios (Behind the Curtain) — "white-collar bloodbath", 2026
- The Wall Street Journal — AI ending the billable hour, 2026
>
💡 Did this article spark an idea for your work?
humanaifit explores how humans and AI truly collaborate. If you face real challenges in enterprise AI adoption, human-AI fit, or global compliance, we'd love to hear from you.
🔗 Search for the "AI Era Survival Guide" (AI时代生存手册) Knowledge Planet, ¥199/year — every article comes with matching toolkits and templates, plus direct discussions with the authors.
💡 What did this article inspire for you?
humanaifit studies how humans and AI can genuinely work together. If you face real questions on enterprise AI adoption, human-AI collaboration, or global compliance, join our discussion.
🔗 Search for the "AI Era Survival Handbook" Knowledge Planet, ¥199/year — every deep article comes with tool templates and direct contact with the author.