Tripling the Ladder -- How IBM Reversed the Most Famous AI Hiring Freeze in Corporate History

Twin Ladder Casebook Series | Twin Ladder | March 2026

The Hook

In May 2023, Arvind Krishna sat for an interview with Bloomberg and said the thing that every CEO was thinking but few would say aloud. IBM, he announced, would pause hiring for back-office roles that artificial intelligence could perform. The number he cited was specific: roughly 7,800 positions -- 30 percent of IBM's 26,000 non-customer-facing workers -- could be replaced by AI and automation over a five-year period. Natural attrition would do the work. The company would simply stop refilling the positions that people left.

The statement landed like a controlled detonation. It was the first time a chief executive of a Fortune 500 company had publicly quantified, by headcount, how many jobs he expected AI to eliminate from his own workforce. The Washington Post ran the headline under the framing "Will AI Take Over Jobs? Answer Is Yes at IBM." Bloomberg, Reuters, CNBC, and Al Jazeera followed within hours. The message rippled through every boardroom in the technology industry and beyond: IBM was not speculating about AI displacement. It was implementing it.

Thirty-three months later, in February 2026, IBM's chief human resources officer, Nickle LaMoreaux, walked onto the stage at Charter's Leading with AI Summit in New York and announced that the company was tripling its entry-level hiring in the United States. "We are tripling our entry-level hiring," she said, "and yes, that is for software developers and all these jobs we're being told AI can do."

The distance between those two statements is not a contradiction. It is a correction. And the story of how IBM traveled from one to the other is the most important case study in the emerging field of AI workforce strategy -- because IBM is the first major technology company to articulate, publicly and precisely, why cutting entry-level hiring in the name of AI efficiency is a false economy that will hollow out the organizations that pursue it.

The Story

The Promise

When Krishna made his May 2023 announcement, the logic appeared unassailable. IBM's back-office functions -- human resources, finance, administrative operations -- were staffed by approximately 26,000 people performing tasks that followed predictable patterns: processing employment verifications, managing benefits inquiries, handling internal HR requests, generating routine reports. These were precisely the categories of work that large language models and automation tools were designed to handle. The arithmetic was straightforward: if AI could absorb 30 percent of this workload, the company could redirect billions in labour costs toward growth investments in cloud computing and enterprise AI -- the divisions that Krishna had staked IBM's future upon.

The timing was deliberate. ChatGPT had launched five months earlier, in November 2022, and the technology industry was in the grip of what would later be recognized as peak AI replacement euphoria. Every earnings call included references to AI-driven efficiency. Every corporate strategy deck featured headcount optimization charts. Krishna's announcement was distinguished not by its ambition but by its candour. Where other executives couched their plans in language about "augmentation" and "workforce evolution," Krishna gave a number: 7,800 jobs. He told Bloomberg directly: "I could easily see 30 percent of that getting replaced by AI and automation over a five-year period."

The market rewarded the clarity. IBM's stock rose on the announcement. Analysts praised the strategic focus. The message to the industry was received: IBM was leading the AI transformation of its own workforce, and the transformation meant fewer people.

The Execution

IBM moved aggressively through 2023, 2024, and into 2025. The company redesigned more than seventy internal workflows through automation and AI. Hiring in back-office functions slowed substantially. Between September 2024 and March 2025, IBM reduced its workforce by an estimated 13,000 to 17,000 employees across two major restructuring waves -- 8,000 to 10,000 in September 2024, followed by 5,000 to 7,000 in March 2025. A further reduction in the fourth quarter of 2025 affected a low single-digit percentage of the global workforce.

The financial results, viewed in isolation, told a success story. IBM's annual revenue for 2025 reached $67.5 billion, a 7.6 percent increase from 2024. The company was growing. It was leaner. The AI thesis appeared to be working.

But inside the organization, a different picture was emerging -- one that the revenue numbers could not capture and the headcount metrics were designed to obscure.

The Reckoning

The problem was not that AI could not perform back-office tasks. It could, and in many cases it performed them faster and more consistently than humans. The problem was what happened to the organization when the humans who had previously performed those tasks were no longer there.

Entry-level employees do not merely execute tasks. They learn the business from the ground up. They develop institutional knowledge -- the understanding of why processes exist, not just how they work. They build the relationships, the contextual awareness, and the professional judgment that organizations require at every level above entry. They are, in the language of workforce planning, the seed corn of the talent pipeline.

When IBM paused hiring for back-office roles, the immediate effect was cost savings. The medium-term effect was a growing gap in the pipeline. The company still needed mid-level managers. It still needed senior professionals with deep institutional knowledge. It still needed people who understood the specific ways that IBM's processes, culture, and customer relationships differed from every other technology company's. Those people had always come from the bottom of the ladder. And the bottom of the ladder was now empty.

IBM's CHRO would later describe this dynamic with precision: reducing junior headcount and entry-level hiring might drive cost savings in the near term, but it risks creating a longer-term scarcity of mid-level managers and experienced workers within the organization. Without the ability to develop their own experienced employees, companies would be forced to look outward in a more costly search for professionalism and expertise. And outside hires, she noted, tend to take longer to adapt to internal systems and culture.

The insight was not novel. It was a rediscovery. Every organisation that has ever built a successful professional development pipeline -- from consulting firms to military services to teaching hospitals -- has understood that you cannot produce senior talent without investing in junior talent. The entry-level role is not an expense to be minimized. It is an investment to be structured. AI had made it possible to automate the tasks that entry-level workers performed. It had not made it possible to automate the learning that performing those tasks produced.

The Reversal

In February 2026, Nickle LaMoreaux made the announcement that placed IBM on the opposite side of its own 2023 strategy. The company would triple entry-level hiring in the United States. The roles would include software developers -- precisely the category that every AI vendor in the market was claiming to be on the verge of automating. The hiring would target Gen Z workers. And the rationale was not charity, employer branding, or public relations. It was strategic necessity.

"The entry-level jobs that you had two to three years ago, AI can do most of them," LaMoreaux explained at the Leading with AI Summit. "So, if you're going to convince your business leaders that you need to make this investment, then you need to be able to show the real value these individuals can bring now. And that has to be through totally different jobs."

The statement contained two critical admissions. First, that IBM's own experience had demonstrated that entry-level roles as previously designed were indeed vulnerable to AI automation. Second, that the correct response was not elimination but redesign -- creating new entry-level positions that were structured around the capabilities that AI could not replicate.

IBM's redesigned entry-level roles look fundamentally different from their predecessors. Junior software developers now spend less time writing routine code -- which AI tools can generate -- and more time working directly with customers to define requirements, understand business context, and translate human needs into technical specifications. New HR hires work not as query processors but as refiners of AI chatbot responses, improving the quality of automated interactions through human judgment and contextual understanding. The roles have been restructured around the one thing that AI cannot do: develop the professional judgment that comes from working with real people on real problems.

LaMoreaux's summary was the sharpest articulation of the case for entry-level investment that any corporate leader has delivered in the AI era: "The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment."

The statement was a prediction. It was also a bet -- one that IBM was backing with headcount, budget, and the professional reputation of its CHRO.

Through the Twin Ladder Lens

IBM's journey from hiring freeze to hiring surge maps directly onto the Twin Ladder framework, and it reveals a pattern that the framework was designed to predict.

The Level 0 Failure: Literacy Without Strategy

When Krishna announced the 7,800-job figure in May 2023, IBM was operating at Level 0 of AI maturity -- AI Literacy. The company understood what AI could do. It could identify the tasks that pattern-matching algorithms could perform more efficiently than humans. It could quantify the cost savings. What it had not yet developed was the capacity to distinguish between tasks that AI could replace and roles that the organization could not afford to lose.

This is the fundamental error at Level 0. AI Literacy enables you to evaluate what AI produces. It does not, by itself, enable you to evaluate what removing humans from the process will cost. The literacy was genuine -- IBM understood the technology better than most. The strategy was incomplete -- because understanding what AI can do is not the same as understanding what your organization needs humans to become.

The Level 1 Gap: No Professional Twin

The Twin Ladder's Level 1 -- the Professional Twin -- exists to prevent exactly the mistake IBM made. At Level 1, an organisation mirrors each professional role with an AI counterpart, not to replace the professional but to create the conditions for comparison, learning, and judgment-building. The human and the AI perform the same work. The human evaluates the AI's output. The AI handles the volume. The human develops the expertise to know when the AI is wrong.

IBM skipped this level entirely in its initial approach. It did not create Professional Twins for its back-office workers. It did not use AI to augment their capabilities while they developed deeper expertise. It removed them from the process. The result was an organisation with powerful AI tools and a diminishing supply of humans who understood the context in which those tools operated.

The February 2026 reversal is, in essence, a return to Level 1 -- but with the roles redesigned. IBM's new entry-level positions are structured as Professional Twin roles, even if the company does not use that terminology. Junior developers work alongside AI coding tools, not as passive reviewers but as active participants who bring customer context that the AI lacks. HR hires refine chatbot outputs, adding the judgment layer that the automated system cannot generate on its own. The humans are not doing the old jobs. They are doing the jobs that emerge when AI handles the routine and humans handle the meaning.

The Level 2 Warning: Operational Twins Require Human Operators

IBM's experience also illuminates why Level 2 -- the Operational Twin -- cannot function without the human foundation built at Level 1. An Operational Twin is a digital replica of a business function, used to test changes before deploying them. IBM has built operational replicas of more than seventy internal workflows. But an operational replica is only as good as the humans who interpret its output, challenge its assumptions, and catch its errors. Without a pipeline of professionals who have climbed from Level 1 -- who understand the business deeply enough to evaluate the model -- the Operational Twin becomes an unaudited system running on inherited assumptions.

This is the deeper lesson of IBM's reversal. The company did not triple entry-level hiring because AI had failed. It tripled entry-level hiring because AI had succeeded at the tasks and the organisation had discovered that tasks are not the same as capabilities. You can automate the task of writing code. You cannot automate the capability of understanding why a customer needs the code written differently. You can automate the task of answering an HR query. You cannot automate the capability of recognising when the query signals a deeper organisational problem.

The Broader Pattern

IBM's reversal is not an isolated event. It is the most prominent data point in an accelerating pattern of AI replacement regret that now spans industries, geographies, and company sizes.

The Data

In February 2026, Gartner published its prediction that by 2027, 50 percent of companies that reduced customer service headcount due to AI would rehire staff to perform similar functions -- often under different job titles such as "Solution Consultants" or "Trusted Advisors." The prediction was grounded in a survey of 321 customer service leaders conducted in October 2025, which found that only 20 percent had actually reduced agent staffing due to AI. Those that did had begun to encounter what Gartner analyst Kathy Ross described as the reality that "as organizations encounter the limits of AI and rising customer expectations, they will need to reinvest in human talent to sustain service quality and growth."

Forrester's Predictions 2026 report quantified the regret with a blunt finding: 55 percent of employers who laid off workers in the name of AI now report regretting the decision. Forrester also identified a cynical undertone to the rehiring trend -- predicting that half of AI-attributed layoffs would be reversed, but that the rehired workers would often be offshore or at significantly lower salaries. The AI narrative, in Forrester's analysis, had become a mechanism for cost restructuring dressed in the language of technological progress.

A survey conducted by an HR analytics firm found that the reversals were happening rapidly: 32.7 percent of companies that conducted AI-led layoffs had already rehired for 25 to 50 percent of the roles they initially eliminated. Another 35.6 percent had rehired for more than half. Over half of HR leaders reported that the rehiring occurred within six months of the original cuts.

Harvard Business Review published an analysis in January 2026, authored by Thomas H. Davenport and Laks Srinivasan, that identified the mechanism driving premature cuts: 60 percent of organisations had already reduced headcount in anticipation of AI's future impact. Not in response to AI's current performance. In anticipation. Companies were firing people based on what AI might do, not what it had demonstrated it could do. HBR noted that attributing layoffs to AI "conveys a more positive message to investors" than citing weak demand or past overhiring -- a finding corroborated by Oxford Economics, which published a research briefing in January 2026 calling AI-attributed layoffs a "corporate fiction" masking routine business cycle adjustments.

The Cases

Klarna remains the most dramatic illustration. The Swedish fintech replaced approximately 700 customer service agents with an OpenAI-powered chatbot in 2024, celebrating two-minute resolution times and $40 million in projected savings. By May 2025, CEO Sebastian Siemiatkowski admitted to Bloomberg that "cost unfortunately seems to have been a too predominant evaluation factor" and that the result was "lower quality." The company began rehiring human agents using an "Uber-style" remote workforce model. (See the Twin Ladder Casebook entry: "We Went Too Far" -- Klarna and the Cost of Replacing Human Judgment.)

Duolingo experienced a compressed version of the same cycle. In April 2025, CEO Luis von Ahn posted a LinkedIn memo announcing the company was becoming "AI-first," including replacing contractors with AI and conditioning new hires on teams' inability to automate their work. The public backlash was immediate and severe -- subscribers threatened cancellations, public commentators accused the company of prioritizing technology over people. Within a week, von Ahn issued a retraction: "To be clear: I do not see AI as replacing what our employees do -- we are in fact continuing to hire at the same speed as before." He later admitted to the Financial Times that he "did not expect the amount of blowback" and that his original memo "did not give enough context."

McDonald's ended its AI drive-through partnership with IBM in June 2024 after a pilot across more than 100 U.S. locations produced accuracy rates in the low-to-mid 80 percent range. Social media documented the failures with viral enthusiasm -- orders with unwanted items, mixed-up lane requests, ignored corrections. The system was fast. It was also wrong often enough to damage the customer experience. McDonald's returned to human order-taking while stating it would seek alternative AI vendors. The pilot now resides in the Museum of Failure's permanent exhibition.

The Commonwealth Bank of Australia walked back a decision to lay off employees following the introduction of a new voice bot system, after discovering that the automated system could not handle the complexity of real customer interactions in financial services.

The Technical Debt Dimension

The pattern extends beyond customer-facing roles into software development, where the consequences of eliminating entry-level workers are compounding into a technical debt crisis. Entry-level tech hiring collapsed by 73.4 percent year-over-year in early 2026, according to industry analysis -- far outpacing the overall hiring decline of approximately 7 percent across seniority levels. The 15 largest technology firms reduced entry-level hiring by 25 percent between 2023 and 2024 alone.

The downstream effects are now measurable. A 2025 Veracode report found that 45 percent of AI-generated code contains OWASP top 10 vulnerabilities. In Java, the security failure rate exceeds 72 percent. CAS Software's analysis of 10 billion lines of code estimated that the world's current technical debt would require 61 billion work days to remediate, driven in part by a fourfold surge in code cloning -- AI generating functionally identical blocks instead of elegant, reusable logic. Stack Overflow documented the phenomenon in January 2026 under the headline: "AI Can 10x Developers ... In Creating Tech Debt."

The industry term emerging for this phenomenon is the "slop layer" -- code that works but nobody understands why, and nobody can fix when it breaks. Without junior developers learning to write code from the ground up, developing the understanding of systems architecture that comes only from building and failing and rebuilding, the slop layer grows. Without a pipeline of developers who have climbed from junior to mid-level to senior, the organisation loses the capacity to audit its own automated output. By 2030, industry observers project a catastrophic shortage of true senior engineers -- those capable of understanding the system below the AI abstraction layer -- precisely because the pipeline that produces them was cut in 2023-2025.

This is the junior developer death spiral. Eliminate junior roles because AI can write the code. Lose the pipeline that produces seniors who can evaluate the code. Accumulate technical debt that nobody remaining has the skills to identify, let alone resolve. Discover, three to five years later, that the cost savings from eliminating junior salaries are dwarfed by the cost of maintaining a codebase that nobody fully understands.

The Korn Ferry Warning

In its 2026 Talent Acquisition Trends report, Korn Ferry surveyed 1,674 global talent leaders and found that 37 percent of organisations plan to replace entry-level roles with AI. At the same time, 73 percent of those same talent leaders identified critical thinking and problem-solving as their number-one recruiting priority -- ranking it above AI technical skills, which placed fifth.

The tension is evident: organisations want employees who can think critically and solve problems, while simultaneously eliminating the roles where critical thinking and problem-solving are developed. Entry-level positions are where professionals learn to navigate ambiguity, build judgment through repeated exposure to real-world complexity, and develop the institutional knowledge that makes them valuable at senior levels. Automating those positions does not eliminate the need for the capabilities they produce. It eliminates the mechanism by which those capabilities are produced.

Korn Ferry's analysts issued an explicit warning: eliminating early-career talent could ultimately choke the pipeline for future leaders. Only 11 percent of talent acquisition leaders said their executives were well-prepared to lead through the AI transition.

Implications for Organisations

IBM's journey from hiring freeze to hiring surge offers five lessons for any organisation navigating the AI replacement question.

First, distinguish between tasks and capabilities. AI can perform tasks. It cannot develop the capabilities that emerge from performing tasks over time within a specific organisational context. The task of writing routine code is automatable. The capability of understanding a customer's business well enough to know which code to write is not.

Second, the talent pipeline is infrastructure, not expense. Entry-level hiring is not a cost to be minimized in the current quarter. It is an investment in the organisation's capacity to function in five years. IBM's CHRO framed this with unusual clarity: the companies that will be most successful in three to five years are those that doubled down on entry-level hiring now, not those that cut it.

Third, redesign roles before you eliminate them. IBM did not simply resume the entry-level hiring it had paused. It redesigned the roles entirely, structuring them around the capabilities that AI cannot replicate -- customer interaction, contextual judgment, quality refinement of AI outputs. The correct response to AI automating an entry-level task is not to eliminate the entry-level role. It is to rebuild the role around the next layer of human value.

Fourth, measure what matters, not what is easy. The companies that cut entry-level workers measured cost savings and task completion rates. They did not measure pipeline health, institutional knowledge retention, or the quality of judgment in complex situations. Klarna measured resolution time and missed trust erosion. IBM measured headcount reduction and missed pipeline depletion. The metrics you choose to track determine the mistakes you will be able to detect.

Fifth, the rehiring will be more expensive than the retention. Forrester and Gartner agree: the companies that cut workers for AI will rehire them -- often within months, frequently at different pay structures, and always with the organisational disruption of rebuilding what was lost. The cheapest way to have mid-level talent in 2029 is to hire and develop entry-level talent in 2026. The most expensive way is to discover in 2029 that you need to acquire it externally, at premium rates, from competitors who invested while you optimized.

Monday Morning Question: If you paused all entry-level hiring today and AI handled the routine tasks perfectly, who in your organisation would be developing the judgment, the institutional knowledge, and the customer relationships you will need your leaders to possess in 2031?

Sources

Bloomberg -- "IBM to Pause Hiring for Jobs That AI Could Do" (May 2023): https://www.bloomberg.com/news/articles/2023-05-01/ibm-to-pause-hiring-for-back-office-jobs-that-ai-could-kill
Fortune -- "IBM Is Tripling the Number of Gen Z Entry-Level Jobs After Finding the Limits of AI Adoption" (February 2026): https://fortune.com/2026/02/13/tech-giant-ibm-tripling-gen-z-entry-level-hiring-according-to-chro-rewriting-jobs-ai-era/
TechCrunch -- "IBM Will Hire Your Entry-Level Talent in the Age of AI" (February 2026): https://techcrunch.com/2026/02/12/ibm-will-hire-your-entry-level-talent-in-the-age-of-ai/
Tom's Hardware -- "IBM Triples Entry-Level Hires for 2026 Despite AI Adoption" (February 2026): https://www.tomshardware.com/tech-industry/artificial-intelligence/ibm-triples-entry-level-hires-for-2026-despite-ai-adoption-bucking-industry-trends-chief-hr-officer-says-that-ai-can-do-most-entry-level-jobs-but-work-still-requires-a-human-touch
Charter -- "Why IBM Is 'Tripling' Entry-Level Hiring as AI Reshapes Jobs" (February 2026): https://www.charterworks.com/ibms-nine-year-ai-journey/
AllWork -- "IBM To Triple Entry-Level Hiring, Warns AI-Driven Hiring Cuts Could Hollow Out Future Leadership" (February 2026): https://allwork.space/2026/02/ibm-to-triple-entry-level-hiring-warns-ai-driven-hiring-cuts-could-hollow-out-future-leadership/
CIO -- "IBM Looks Beyond Short-Term AI Gains, Tripling Entry-Level Hiring" (February 2026): https://www.cio.com/article/4134276/ibm-looks-beyond-short-term-ai-gains-tripling-entry-level-hiring.html
Gartner -- "Gartner Predicts Half of Companies That Cut Customer Service Staff Due to AI Will Rehire by 2027" (February 2026): https://www.gartner.com/en/newsroom/press-releases/2026-02-03-gartner-predicts-half-of-companies-that-cut-customer-service-staff-due-to-ai-will-rehire-by-2027
Forrester -- "Predictions 2026: The Future of Work" -- cited via The Register (October 2025): https://www.theregister.com/2025/10/29/forrester_ai_rehiring/
Harvard Business Review -- "Companies Are Laying Off Workers Because of AI's Potential -- Not Its Performance" (January 2026): https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance
Fortune -- "AI Layoffs Are Looking More and More Like Corporate Fiction" -- Oxford Economics (January 2026): https://fortune.com/2026/01/07/ai-layoffs-convenient-corporate-fiction-true-false-oxford-economics-productivity/
Korn Ferry -- "Korn Ferry Research Unveils Top Talent Acquisition Trends Shaping 2026": https://www.kornferry.com/about-us/press/korn-ferry-research-unveils-top-talent-acquisition-trends-shaping-2026
Inc. -- "A New Report Says AI Layoffs Are Backfiring and Half of Companies Will Start Rehiring" (2026): https://www.inc.com/bruce-crumley/new-report-says-ai-layoffs-are-backfiring-and-half-of-companies-will-start-rehiring/91297210
CNBC -- "McDonald's to End AI Drive-Thru Test with IBM" (June 2024): https://www.cnbc.com/2024/06/17/mcdonalds-to-end-ibm-ai-drive-thru-test.html
Fortune -- "Duolingo CEO Walks Back AI-First Comments" (May 2025): https://fortune.com/2025/05/24/duolingo-ai-first-employees-ceo-luis-von-ahn/
WhatJobs -- "The AI Developer Replacement Plan Has Failed: Technical Debt, Security Flaws, and the Junior Death Spiral": https://www.whatjobs.com/news/the-ai-developer-replacement-plan-has-failed-technical-debt-security-flaws-and-the-junior-death-spiral/
Stack Overflow -- "AI Can 10x Developers ... In Creating Tech Debt" (January 2026): https://stackoverflow.blog/2026/01/23/ai-can-10x-developers-in-creating-tech-debt
Veracode -- Gen AI Security Report (2025) -- cited via WhatJobs and Quasa.io
HR Dive -- "If AI Kills the Entry-Level Job, Employers May Not Be Ready for What Comes Next": https://www.hrdive.com/news/AI-entry-level-jobs-talent-pipeline/809413/
Diginomica -- "IBM's CHRO Preaches Fluent AI Jobs Blasphemy": https://diginomica.com/ibms-chro-preaches-fluent-ai-jobs-blasphemy-heres-why-its-critical-business-leaders-everywhere-pay