> ## Content Index
> Fetch the complete content index at: https://www.groktop.us/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Most Human AI Strategy May Be Moving People, Not Replacing Them
- URL: https://www.groktop.us/moving-people-not-replacing-them/
- Published: 2026-09-10T15:44:13.000Z
- Updated: 2026-09-10T15:44:12.000Z
- Description: The most human AI strategy treats automation as a reason to move people into better work, not discard them.
- Author: Magnus Hedemark

The easiest AI strategy fits on one slide: automate a task, delete a role, and claim the savings. The math looks clean. The company underneath it usually isn't.

Work doesn't disappear that cleanly. It lands on somebody else's desk. Decision rights shift, the old bottleneck turns up in a new department, and the spreadsheet stops being useful. The people who know how the business really works are still the shortest path from a clever system to something customers and coworkers can use.

This is chapter five of the argument I've been building across this series. The first four chapters dealt with the usual shortcuts: buying a tool and calling it strategy, treating headcount reduction as the prize, or handing everyone a course and hoping the organization sorts itself out. Now comes the practical question. When AI changes the work, where do the people who know that work go next?

I don't believe every job can or should survive unchanged. Some roles will shrink or disappear. But many will recombine: responsibilities once divided among specialists will overlap and collect around broader roles, as the first article in this series argued. The mistake is treating displacement as the first move. When AI takes on part of a role, leaders should look for the adjacent skills, real openings, and paid bridge that can carry the person into growing work.

That's not charity. It's how you keep hard-won judgment inside the company while the work is being rebuilt. It also answers [the AI hiring reversal](https://www.groktop.us/the-ai-hiring-reversal-why-headcount-reduction-was-always-the-wrong-goal/): more capability is the prize, not the neatest possible payroll.

## The useful signal is movement, not applause

Companies love an announcement: an academy, a billion-dollar training commitment, fewer degree requirements, or a pledge to teach everybody AI. Any of that may help. None of it answers the question I care about: did anybody reach better work?

[LinkedIn's 2026 Top Companies methodology](https://news.linkedin.com/2026/LinkedIn-Top-Companies-2026?ref=groktop.us) gives me a better place to look. Three of its eight pillars are the ability to advance, skills growth, and company stability. Advancement includes promotions and moves to other companies, while skills growth tracks what people add while employed. JPMorgan Chase ranked first, Microsoft third, Walmart seventh, and Bank of America tenth.

That signal comes with a warning label. LinkedIn builds the ranking from member and profile data, covers large employers, and excludes companies that crossed specified attrition or announced-layoff thresholds during the measurement window. It can point us toward practices inside large, comparatively stable companies. It can't tell us that one program caused a promotion, prevented a layoff, or would work across the rest of the economy.

I use the ranking as a map to the machinery, not proof that the machinery works.

## The machinery is beginning to appear

Bank of America gives us the clearest sign that internal movement can become a real staffing channel. LinkedIn says the bank is filling **thousands** of roles internally, and Bank of America describes formal [career-growth and employee-training programs](https://careers.bankofamerica.com/en-us/benefits/career-growth?ref=groktop.us).

But “thousands” is about as far as this source lets us go. We don't have a denominator, a time period, a breakdown of destination roles, or results for pay and retention.

![Automation becomes human-centered when it opens a measured path into new work.](https://storage.ghost.io/c/f1/0e/f10e80f4-9285-43fc-acd4-35910a12c5f0/content/images/2026/09/automation-to-redeployment.png)

Automation becomes human-centered when it opens a measured path into new work.

I can see the mechanism. I can't yet tell whether the workers came out ahead.

JPMorgan Chase and Microsoft show another part of the machinery: broad AI learning connected to everyday work. LinkedIn says both companies are embedding AI while training employees to build and use the tools, including a Microsoft effort to teach every employee how to build AI tools. JPMorgan Chase also describes a portfolio of [career and skills programs](https://www.jpmorganchase.com/impact/careers-and-skills?ref=groktop.us).

This is where big learning programs can turn into corporate wallpaper. An enterprise-wide announcement doesn't tell us who finished, who used the training in a live workflow, whose role changed, or who kept a job. Training becomes mobility infrastructure only when it leads to actual openings and managers are expected to hire from it.

And this is where [organizational infrastructure becomes the real rate limiter](https://www.groktop.us/org-chart-rate-limiter/). A learning platform can't move anybody while job architecture, compensation bands, hiring incentives, and departmental budgets keep that person locked inside a local box.

## Build entry ramps, not just training libraries

Training matters only when there's a door into an actual job. IBM and Accenture show two ways to build that door. LinkedIn reports that IBM plans to triple United States entry-level hiring in 2026\. IBM's documented [skills-centered talent strategy](https://www.linkedin.com/business/talent/blog/talent-acquisition/how-ibm-centered-talent-strategy-on-skills?ref=groktop.us) includes apprenticeships in fields such as cybersecurity and software development. Accenture describes [apprenticeships as an alternative pathway](https://www.accenture.com/us-en/careers/life-at-accenture/apprenticeships?ref=groktop.us) that doesn't depend on the four-year-degree cycle and can help people reskill for new opportunities.

![Training matters only when it leads to a real destination role, authority, and advancement path.](https://storage.ghost.io/c/f1/0e/f10e80f4-9285-43fc-acd4-35910a12c5f0/content/images/2026/09/skills-first-entry-ramp.png)

Training matters only when it leads to a real destination role, authority, and advancement path.

JPMorgan Chase's [Emerging Talent Experience](https://www.jpmorganchase.com/careers/explore-opportunities/programs/et-experience?ref=groktop.us) opens that door beyond one traditional pipeline. It includes undergraduates, people up to three years out of college, and graduates of skills-based workforce programs such as apprenticeships and coding boot camps.

“Learn AI” and “change jobs” are not the same offer. A real transition needs an opening, a selection process that recognizes adjacent ability, supervised practice, and a role waiting on the other side. Miss one of those, and the worker gets a content library while the economic problem stays exactly where it was.

Walmart offers a larger-scale example. The company says [90 percent of its United States roles don't require a college degree](https://corporate.walmart.com/skillsfirst?ref=groktop.us) and frames advancement around what associates can do rather than where they started. Walmart also says it's [investing $1 billion by 2026](https://corporate.walmart.com/news/2025/04/07/creating-opportunity-for-all-american-workers?ref=groktop.us) in training, education, and paths toward jobs with greater responsibility and pay. A Walmart-convened effort described by the [Burning Glass Institute](https://www.burningglassinstitute.org/research/skills-first-phase-2?ref=groktop.us) covers 30 roles encompassing more than 35 million workers.

Those numbers tell us the scale of the promise, not whether it paid off for workers. We still need completion rates, occupational moves, promotions, wage growth, and displacement avoided. A billion-dollar input can end in a weak outcome. Spending isn't a crossing.

## A redeployment operating model

Chapter five has to end with machinery, not another principle. If leaders want movement instead of theater, the workforce decision belongs in the room when the work is redesigned. Waiting until the new organization chart hardens is too late.

I'd build the operating model around five moves:

1. **Map tasks before cutting roles.** Start with what is shrinking, what is becoming more valuable, and which nearby capabilities the people doing the work already have. A role is a bundle of tasks. Don't treat it like one indivisible cost unit.
2. **Open the destination before closing the origin.** Put the growing roles where people can see them, define the bridge skills, and give internal candidates a real shot before starting an external search. A course with nowhere to go isn't a mobility program.
3. **Pay for the crossing.** Protected learning time, paid apprenticeships, supervised practice, and temporary capacity coverage belong in the transformation budget. Asking someone to rebuild a career at night quietly sends the bill to the worker.
4. **Stop rewarding managers for hoarding talent.** Give leaders credit when someone they developed moves into a stronger role elsewhere in the company. If a manager loses budget or status whenever a good person leaves, the system will pin talented people inside the teams least willing to release them.
5. **Make involuntary exit require an explanation.** Not every transition will work. Before an exit, leaders should still be able to show which adjacent paths they considered, offered, and attempted, plus why the crossing failed.

And yes, this is governance. As [AI governance remains human work](https://www.groktop.us/ai-governance-human-work/), no model can decide whether an opportunity was fair, a destination was credible, or a failure rate was acceptable. A training-completion dashboard can't decide that, either. People have to own the judgment.

## Measure whether anyone actually moved

Press releases count seats, courses, and dollars. The executive scoreboard should count crossings.

Track the share of affected workers offered a bridge path and the share who enter one. Then watch training-to-role conversion, median wage change, time to proficiency, 12-month retention, demographic parity, involuntary exits, and net employment. Break the results out by prior role, destination role, business unit, geography, and worker population. A healthy average can hide a lot of people who never got through the door.

![A human-centered strategy measures transitions, pay, retention, and role quality, not training activity alone.](https://storage.ghost.io/c/f1/0e/f10e80f4-9285-43fc-acd4-35910a12c5f0/content/images/2026/09/measure-movement.png)

A human-centered strategy measures transitions, pay, retention, and role quality, not training activity alone.

Then put internal redeployment beside external hiring. If a company says it can't find AI talent while its trained employees can't reach an interview, the skills shortage may not be the bottleneck. The internal market may be broken. If people finish programs but land in lower-paid or fragile work, the company moved bodies without creating mobility. When aggressive automation and growing hiring appear in the same announcement, ask whether the people affected by one had a credible path into the other.

The evidence I have today can't name a winner. It can support a better executive question: how many people can this organization move into more valuable work before displacement becomes necessary?

The work will change. That's why this matters. A serious AI strategy gives the people closest to that work a funded, measured way to change with it. Replacement comes only after the organization has done the harder job of building somewhere real for them to go.