The Two Paths: AI Workforce Strategy — Cut or Cultivate?

1. THE KEY QUESTION WHEN AI IS INTRODUCED

Every major organisation deploying AI today faces the same inflection point: what happens to the people whose work AI can now do? The answer is the single most consequential strategic decision a company will make during its AI transformation.

Two paths exist. The first — workforce reduction — treats AI as a cost-elimination tool. The second — workforce repurposing — treats AI as a capability multiplier. Both paths have been taken by real companies. The outcomes diverge dramatically.

Research now confirms that 55% of organisations that made significant AI-driven layoffs report regretting the decision, with quality degradation, institutional knowledge loss, and talent flight consistently exceeding the savings projected in the original business case. Meanwhile, companies that invested in reskilling and redeployment — led by IKEA's landmark interior design transformation — are generating new revenue streams, building deeper customer relationships, and winning the talent market for the AI era.

Core Finding:

"Layoffs do not deliver AI ROI. Redeployment does." — Gartner Research, 2026

2. THE SCALE OF THE DECISION

2.1 The Macro Workforce Transformation

The World Economic Forum's Future of Jobs Report 2025, drawing on perspectives of over 1,000 leading global employers representing more than 14 million workers, projects that by 2030, approximately 92 million jobs will be displaced by AI and related technologies — while 170 million new roles will be created, yielding a net gain of 78 million jobs. The critical phrase is 'net gain.' The question is not whether AI destroys more jobs than it creates — it does not. The question is whether the workers displaced are the same workers who fill the new roles, and that outcome depends entirely on strategic choices made at the organisational level.

Among employers surveyed, 40% anticipate reducing their workforce by automating tasks with AI, while 77% plan large-scale upskilling and reskilling initiatives. The divergence between these two groups will define competitive positioning for the next decade. AI and information processing will affect 86% of businesses by 2030.

2.2 Why This Decision Matters More Than It Appears

BCG research involving hundreds of companies found that roughly 10% of AI value comes from algorithms, 20% from technology implementation, and the remaining 70% from rethinking the people component. Future-built companies plan to upskill more than 50% of their employees on AI, compared with only 20% for laggards, and are four times more likely to have structured AI-learning programmes. These companies achieve five times the revenue increases and three times the cost reductions compared to peers.

3. PATH ONE — WORKFORCE REDUCTION

3.1 The Logic and the Temptation

When AI can handle a significant share of the existing workload, the arithmetic of headcount reduction becomes immediately attractive. If an AI chatbot handles 47–57% of customer service queries, a CFO who sees 500 customer service roles on the payroll will naturally ask: Can we run this function with 250 people? The business case writes itself on a spreadsheet. Headcount cost eliminated. AI licensing cost added. Net saving calculated. Board presentation prepared.

This is the path taken by a growing list of companies — Klarna, IBM, Salesforce, Duolingo, and many others — and it is one with consistent, well-documented, and predictable consequences.

3.2 Case Study: Klarna — The Most Cited Cautionary Tale

In 2024, Swedish fintech Klarna became the most-discussed example of AI replacing human workers at scale. The company deployed an OpenAI-powered customer service chatbot that, within its first month, was handling 75% of all customer chats — 2.3 million conversations across 23 markets and 35 languages. The CEO froze all hiring for over 12 months and publicly declared that AI could already do 'all of the jobs that we as humans do.' Headcount fell from 7,400 to approximately 3,000.

The story intended to demonstrate AI's supremacy, but it became one of the clearest illustrations of why full replacement fails. Customer satisfaction data deteriorated on complex service interactions. Downstream costs — quality remediation, escalated handling of complex cases, and eventual rehiring — consumed more than was saved. By early 2026, Klarna was quietly rebuilding its customer service capability. The CEO who once wanted to eliminate 700 positions now describes quality human support as 'the way of the future.'

3.3 Case Study: IBM — 8,000 Roles Eliminated

IBM eliminated approximately 8,000 positions, projecting over $640 million in annual savings as part of a broader AI-driven restructuring, citing a shift toward AI in areas including data entry, customer service, and entry-level programming. IBM has simultaneously invested in AI-powered skills intelligence platforms designed to redeploy workers internally, and has reported success in transitioning employees into cloud computing and emerging technology roles when this approach is followed. The tension at IBM illustrates the complexity of the reduction path: when cutting is accompanied by meaningful investment in internal mobility, outcomes improve. When cost-cutting dominates, the pattern of regret repeats.

3.4 Case Study: Salesforce — Cuts and Quiet Redeployment

Salesforce CEO Marc Benioff announced a reduction of the company's customer support function from 9,000 to approximately 5,000 heads, citing AI agents as the enabler. The reality was more nuanced: a Salesforce spokesperson later confirmed that the company had actually 'redeployed hundreds of employees into other areas', including professional services, sales, and customer success — rather than simply eliminating roles. The public framing as AI replacement and the private reality of redeployment illustrate a pattern seen across the industry.

3.5 Case Study: Duolingo — Brand Damage and Contractor Controversy

In January 2024, Duolingo reduced its contractor workforce by approximately 10%, primarily affecting content translators and writers, as the company shifted to GPT-4 for content generation. The controversy accelerated sharply in April 2025 when CEO Luis von Ahn posted an 'AI-first' memo outlining a strategy to gradually reduce contractor roles that AI could handle. The public response was severe — a company that had built its brand on accessible, humanised education became a case study in AI-driven workforce reduction. Von Ahn later admitted the memo 'did not give enough context' and acknowledged that the company had underestimated public reaction.

4. THE HIDDEN COSTS OF WORKFORCE REDUCTION

4.1 What the Spreadsheet Doesn't Capture

The financial models used to justify AI-driven layoffs consistently undercount the costs of the decision. Research data across multiple dimensions paints a consistent picture across six critical categories:

Quality Degradation: 74% of companies that reported regret over AI-driven cuts saw measurable quality degradation in the first year. For customer support functions, this translated to higher escalation rates and declining satisfaction scores that took an average of 14 months to reverse.
Institutional Knowledge Loss: Nearly one-third (32.9%) of HR leaders reported that their organisations lost critical skills and expertise following AI-driven restructuring. A further 28.1% said the remaining workforce lacked the capabilities needed to fill those knowledge gaps. This loss cannot be recovered solely through AI tools.
Rehiring Costs: Among organisations that conducted AI-driven layoffs, 32.7% have already rehired between 25–50% of the roles they eliminated, and 35.6% have brought back more than half. Forrester Research calculated that companies ended up paying 1.27 times the cost of their layoffs due to knowledge gaps and reduced productivity, before the cost of rehiring. Gartner projects that by 2027, half of companies that cut staff for AI will rehire them.
Talent Flight: 81% of companies reporting regret over AI-driven cuts saw elevated voluntary turnover among retained employees in the 12 months following reductions. High-performing employees — those most capable of working alongside AI — are the first to leave. The departure rate among high performers in companies with no AI upskilling investment runs approximately 22% above the industry baseline.
Survivor Syndrome: Harvard Business Review meta-analysis found that after layoffs, companies experienced a 25% decrease in performance and a 31% decline in employee morale among remaining staff. Employees who stay experience heightened anxiety, grief, increased workload without additional support, and a fundamental erosion of trust in leadership.
Partial Automation Reality: Among organisations that pursued AI-driven role elimination, only 21.4% said automation fully replaced roles without operational problems. The majority — 66.1% — said AI successfully replaced only some tasks, not entire jobs. The AI tools that justified cuts handled the easy 80% of cases while failing unpredictably on the 20% that mattered most.

5. PATH TWO — WORKFORCE REPURPOSING

5.2 The Strategic Logic

The repurposing path begins with a different question. Rather than asking 'how many people can we eliminate?' it asks 'what becomes possible with these people now that AI is handling routine work?' This is not an act of charity — it is a business strategy grounded in the recognition that human capability, institutional knowledge, and customer relationships are competitive assets, not line items to be optimised away.

The Gartner analysis is unambiguous: 'Many CEOs resort to layoffs in an attempt to showcase quick returns from AI; however, this strategy is misguided. While workforce reductions may create budget flexibility, they do not yield actual returns. Organisations that successfully improve ROI are those that do not eliminate the need for human workers but rather enhance their roles by investing in skills, responsibilities, and operational models that enable humans to oversee and expand autonomous systems.'

5.3 Case Study: IKEA — The Benchmark for AI Workforce Transformation

IKEA's AI workforce transformation is the most widely referenced example of the repurposing path executed at scale. An AI chatbot named Billie — named after the company's famous Billy bookcase range — was launched in 2021 and began handling approximately 47–57% of all customer queries without human escalation. The conventional response to this automation rate in a call centre would have been to reduce headcount proportionally.

Ingka Group, the largest IKEA franchisee, did not run that calculation. Instead, the company studied the 43–53% of conversations that Billie could not resolve. These unresolved cases revealed a pattern of unmet customer demand: customers wanted design help — how to fit a kitchen into an awkward corner, how to plan a child's room, how to make a small apartment feel bigger.

IKEA identified that its 8,500 call centre workers possessed an asset that could not be replicated from scratch: deep, comprehensive knowledge of the IKEA catalogue. This knowledge was the exact foundation needed to deliver interior design consulting. The company launched a structured reskilling programme, training these employees to become remote interior design advisers who conducted consultations via phone and video.

Within the new service model, AI tools supported the design advisers: generating design concepts, exploring configurations, and speeding up decision-making. The human advisers brought judgment, empathy, contextual understanding, and relationship capability. The AI brought speed and analytical processing. The combination produced a service that neither could deliver on its own.

The IKEA Result

A new service line generating an estimated €1 billion to €1.3 billion in new revenue. IKEA's colleagues are now doing more meaningful work, customer experience has been elevated through more substantive human interaction, and a billion-euro service business has been built using people the company already employed.

5.4 Case Study: JPMorgan Chase — 90% Successful Role Transition

JPMorgan Chase used AI-powered skills mapping to identify employees whose routine tasks were being automated and determine what higher-value activities they could transition into. The outcome is exceptional: over 90% of employees whose original roles were significantly impacted by automation successfully transitioned to new positions within the company. Retention rates for these employees exceeded 85% two years after transition — compared to industry averages, suggesting that major role disruptions typically lead to 40–50% voluntary turnover.

5.5 Case Study: Amazon — US$1.2 Billion Upskilling Commitment

Amazon's 'Upskilling 2025' initiative committed US$1.2 billion over five years to provide tech upskilling to employees. Amazon Web Services independently committed to providing free cloud computing skills training to 29 million people worldwide — a goal achieved ahead of schedule, with more than 31 million learners across 200 countries trained. Specialised programmes including Machine Learning University and a Mechatronics and Robotics programme have enabled employees to move into higher-value technical roles, with the robotics programme delivering hourly wages up to 40% higher upon completion.

5.6 Case Study: Accenture — 550,000 Trained, 77,000 AI Specialists

Accenture has trained more than 550,000 employees in generative AI fundamentals — approximately 70% of its nearly 779,000-person workforce. The company nearly doubled its AI and data specialist cohort to 77,000 within two years. The company recorded US$5.9 billion in generative AI bookings for fiscal 2025. The investment in people is directly connected to the investment in client revenue.

6. THE FINANCIAL CASE FOR REPURPOSING

6.1 Upskilling vs. Replacement: The Numbers

Internal reskilling is consistently more cost-effective than external replacement. Research indicates that utilising an AI skills intelligence system to identify and reskill an existing employee is approximately 23% more cost-effective than hiring externally. External hires for AI-fluent roles command salary premiums of 15–25% above their non-AI-fluent equivalents, and recruiting fees and ramp-up time add further cost.

The total per-employee cost of upskilling an existing worker ranges from approximately US$10,600 to US$27,400 — compared to external hiring costs that run 1.5 to 2x annual salary for mid-market professional roles when turnover, recruiting, and onboarding are fully accounted for.

Deloitte's 2025 Human Capital Trends found that organisations investing in workforce development were 1.8 times more likely to report better financial results. BCG research shows that future-built companies achieve five times the revenue increases and three times the cost reductions from AI compared to laggards.

7. CHOOSING THE RIGHT PATH — A STRATEGIC FRAMEWORK

7.1 The Four Diagnostic Questions

Before making any workforce decision in the context of AI deployment, organisations should work through four diagnostic questions:

Q1 — What is AI actually replacing — tasks or judgment? The evidence consistently shows that AI successfully replaces tasks, not roles. The relevant question is not 'can AI do this job?' but 'what proportion of this job's tasks can AI handle, and what remains?' Most roles contain a portfolio of tasks, some routine and automatable, others requiring contextual judgment, emotional intelligence, and human relationship.
Q2 — What institutional knowledge does this workforce carry? The cost of institutional knowledge loss consistently exceeds projections. Before any reduction decision, organisations must map what knowledge lives in the workforce — product expertise, relationship capital, customer context, process understanding — and determine whether that knowledge can be preserved, transferred, or encoded.
Q3 — Where is unmet customer demand? Every organisation has market signals hidden in its interaction data, its complaint logs, and its escalation patterns. The repurposing path requires the discipline to read those signals and design new service propositions around them. IKEA's transformation was driven entirely by data analysis of what customers were actually asking for that the existing service model could not deliver.
Q4 — What does the talent market look like in 24 months? AI-fluent talent is increasingly scarce and expensive. Organisations that eliminate experienced, company-knowledgeable people today to save labour costs may find themselves competing expensively for AI-competent replacements within 12–24 months. The companies building the strongest competitive positions treat their existing workforces as the foundation for AI-augmented capability.

7.2 The Repurposing Roadmap

For organisations committed to the repurposing path, the following phased sequence reflects best practice across documented case studies:

Phase 1 — AI Deployment & Interaction Analysis (Months 1–6): Deploy AI capability to handle routine, high-volume interactions. Simultaneously, build a rigorous analysis infrastructure to study what AI cannot handle — escalation patterns, unresolved queries, and interactions that require human judgment. This data becomes the foundation for service redesign.
Phase 2 — Skills Mapping & Gap Analysis (Months 3–6): Map the existing workforce's skills, knowledge, and aptitudes against the roles and capabilities identified in the interaction analysis as needed. Use AI-powered skills intelligence to identify 'adjacent skills' — capabilities that provide a bridge from a declining role to an emerging one.
Phase 3 — Structured Reskilling Programme (Months 6–18): Design and execute targeted training that builds from existing knowledge rather than starting from scratch. The most effective programmes are built around actual workflow change — embedding AI tools into the new roles as learning occurs, rather than certifying theoretical knowledge.
Phase 4 — New Service Launch & Iteration (Months 12–24): Launch the new service proposition with the repurposed workforce, supported by AI tools that augment their capability. Measure outcomes — revenue generated, customer satisfaction, employee engagement — and iterate based on results.

8. THE HUMAN DIMENSION

8.1 The Culture Signal AI Deployment Sends

Every workforce decision made during an AI deployment sends a signal to every employee: this is what we value, and this is what you can expect. The organisations that treat AI as a headcount reduction tool send a signal that human workers are temporary — cost items to be replaced as soon as technology permits. The talent paradox that results is not a side effect. It is the logical response of rational people to accurate information.

The organisations that treat AI as a capability multiplier send a different signal: we are investing in your development because we believe the combination of your knowledge and AI capability creates more value than either alone. This signal produces the opposite talent dynamic — retention of experienced people, attraction of AI-fluent external talent, and the psychological safety needed for employees to engage with AI tools honestly rather than defensively.

8.2 The Ethical Case as a Business Case

Beyond strategy and economics, the workforce repurposing path reflects a genuine ethical commitment increasingly visible to customers, partners, and regulators. Duolingo's 'AI-first' memo generated consumer backlash because it violated an implicit contract: the expectation that a company committed to human learning would treat its own human workers with corresponding care.

IKEA's transformation has become a case study precisely because it demonstrates that the ethical choice — investing in people — and the strategic choice — building a billion-euro service business — were one and the same. Deloitte's research confirms this alignment: organisations that invest in workforce development are 1.8 times more likely to report better financial results. The ethical and the economically rational converge on the same conclusion

9. The Path for Growth

The data, the case studies, and the emerging research consensus point in the same direction. AI workforce transformation is not a choice between human workers and technology. It is a choice about what kind of organisation you are building.

Path One — workforce reduction — offers the illusion of an immediate return: a visible headcount decline, a savings figure in the board presentation, and a narrative about AI efficiency for investor calls. The actual outcomes, documented across hundreds of companies, are declining quality, departing talent, institutional knowledge gaps, expensive rehiring, and net financial results that frequently fall below baseline.
Path Two — workforce repurposing — requires more from leadership. It demands analytical rigour in understanding what AI cannot do, creative thinking about what unmet customer needs remain, investment in structured reskilling, and patience through a transition period. The outcomes, documented in IKEA's €1+ billion revenue line, JPMorgan's 90% successful transition rate, and Amazon's 31-million-person training ecosystem, are a durable competitive advantage, workforce capability that compounds over time, and organisations positioned to capture the 170 million new jobs the WEF projects will be created by 2030.

The Final Word

IKEA looked at 8,500 people and saw a billion-euro opportunity. The companies that followed the other path looked at the same situation and saw a cost to cut. The results are in. The question is not whether AI will transform your workforce — it already is. The question is whether the humans in your organisation will be the ones who build what comes next.