The Conversation: Five Years Later

Issue 003 · 10 April 2026

In 2020 and 2021, the world's most cited thought leaders in HR and workforce strategy made their boldest predictions. They had global platforms, enormous research budgets, and decades of credibility. Five years later, the evidence is in.

---

The verdict at a glance

| Leader | Prediction period | Score | Biggest miss | |---|---|---|---| | World Economic Forum | 2020 Future of Jobs | D+ | Wrong mechanism, wrong timeline, wrong geography | | Josh Bersin | 2021-2024 Annual Predictions | C+ | Talent war reversed; psychological cost absent | | Deloitte Human Capital Trends | 2021-2026 Annual Report | C+ | Global South absent; speed underestimated | | Dave Ulrich | 2020-2021 HR principles | B- | Principles held; assumptions about speed did not |

---

World Economic Forum - Grade: D+

The prediction (October 2020): 85 million jobs displaced by 2025. 97 million new roles created. Net gain 12 million. "The robot revolution will create more than it destroys."

The evidence:

The WEF's own 2025 report quietly moved the goalposts. The 2020 5-year prediction became a 2025 10-year projection: 92 million displaced, 170 million created, by 2030. The specific 85 million displacement figure was never validated.

The mechanism was wrong. The 2020 report predicted displacement of manufacturing and routine physical roles. What actually happened: 30% drop in entry-level analytical and knowledge work postings since 2022. The wrong jobs disappeared first.

The geography was absent. India, GCC, and Africa received two index mentions across 100 pages built on surveys of 26 economies. The regions with the highest AI exposure got the least analysis.

The speed was wrong. "Work will be divided equally between humans and machines by 2025" - Oracle sent 30,000 termination emails before breakfast on a Tuesday in March 2026. The company had just posted a 95% jump in net income.

What they got right: Direction. AI disruption was real. Reskilling urgency was real. Everything else - timing, mechanism, geography, scale - missed.

---

Josh Bersin - Grade: C+

The prediction (2021-2024): New "war for talent" driven by demographic decline. Skills-based organisations rising. AI coaches within "a year or two." AI would "superpower" every employee.

The evidence:

Within 24 months of his "talent war" predictions, Bersin was writing "Yes, AI is really impacting the job market." The war ended. 21% of companies have already stopped hiring entry-level employees. One in three expects to eliminate those roles entirely by end of 2026.

The Superworker thesis is inspiring but incomplete. University of Florida researchers have proposed a clinical framework called AI Replacement Dysfunction (AIRD) to describe precisely what the Superworker model left out: the psychological cost of AI transition anxiety. Spring Health research found that frequent AI users experience 45% higher burnout rates than non-users. The framework empowered workers without accounting for the cost of constant adaptation.

Skills-based organisations: managers still spend 40% of their time on administrative tasks and only 13% developing people. 36% of managers say they are insufficiently prepared to be people managers. The transformation has not materialised at the pace predicted.

What he got right: AI domination of the HR tech stack - correct. Productivity as the core metric - correct. Learning as a strategic variable - correct. The direction was right. The human cost and geographic scope were wrong.

---

Deloitte Human Capital Trends - Grade: C+

The prediction (2021-2026): Four possible futures for the worker-employer relationship. Boundaryless organisations. Workforce ecosystems. Human-AI collaboration as gradual co-evolution.

The evidence:

Deloitte's own 2026 report: only 6% of leaders say they are making progress designing human-AI interactions. 65% of organisations believe their culture needs to change significantly because of AI. The "gradual co-evolution" frame did not survive Oracle's 6am email.

The boundaryless organisation prediction was directionally correct but the boundary that dissolved first was not the hierarchy - it was the employer-employee trust relationship. Cloudflare's 2026 Threat Report documents organised laptop farm operations exploiting exactly the boundary-free work environment Deloitte celebrated.

One-third of workers experienced 15 major changes last year alone. Only 27% of leaders say their organisations manage change well. Deloitte's own 2026 data indicts the optimism of their earlier editions.

What they got right: Workforce ecosystem framing - prescient. Skills-based direction - correct. The 2026 "tipping points" language shows genuine learning from the pace mismatch. But 11 consecutive annual reports with structural absence of the Global South is not a minor omission.

---

Dave Ulrich - Grade: B-

The prediction (2020-2021): Organisation trumps individual talent (4x impact). Work boundaries shift from place to values. HR must create value for external stakeholders. Harness uncertainty as a capability.

The evidence:

The values-not-place prediction was validated - but not as intended. Deepfake job candidates, North Korean laptop farm operations, and AI-generated CVs exploiting the boundary-free work environment his framework described approvingly. The boundary dissolved. So did identity verification.

The 4x organisation claim remains structurally sound. But the University of Florida's AIRD framework shows what happens when the individual is not supported through the transition - psychological collapse becomes an organisational capability problem. The boundary between individual wellbeing and organisational performance is thinner than the model assumed.

What he got right: External stakeholder orientation - correct and ahead of its time. Harnessing uncertainty as a capability - correct and more urgent in 2026 than in 2021. The principles held. The speed and mechanism did not fit the framework.

---

The four blind spots all four share

1. The Global South was invisible. India, GCC, Africa - 2 billion workers, the highest AI exposure, absent from every major prediction framework. India's silent hiring freeze. The GCC nationalisation-automation collision. Africa's 2,000 languages and 36% connectivity gap. None modelled.

2. The psychology was absent. University of Florida researchers have proposed AIRD - AI Replacement Dysfunction - as a clinical framework for AI transition anxiety. Spring Health research found 45% higher burnout among frequent AI users. Moodle data shows 13% cite AI worry as their primary burnout driver. Five years of predictions about AI and the future of work said virtually nothing about what happens inside the human being going through it.

3. The entry-level pipeline collapse was not modelled. Every framework assumed the career ladder stayed intact. Entry-level postings down 30% since 2022. 21% of companies have stopped junior hiring. One in three expects elimination of entry-level roles by end of 2026. The rungs are gone. No framework predicted this.

4. The speed was wrong by an order of magnitude. Every prediction assumed 3-5 year transition windows. Oracle moved in one morning. Gemma 4 - frontier AI on a smartphone in 140 languages, offline, free - arrived 18 months ahead of any prediction schedule.

---

The reAImagine.work prediction check - Issue 001 vs now

For transparency, here is our own six-week audit.

AI Output Evaluation: Predicted peak demand in 6-12 months. Revised to 3-6 months. Physical AI has broadened the skill beyond digital outputs. Underestimated scope.

African-Language NLP: Predicted 18-36 months. Revised to 9-18 months. Gemma 4 offline multilingual capability compressed the timeline significantly. Underestimated pace.

Transactional Customer Service depreciation: Confirmed and accelerating. Gemma 4 offline adds displacement even in low-connectivity markets.

Document-Based Research and Synthesis depreciation: Confirmed. Junior hiring decline has continued.

We got two predictions approximately right. We underestimated two. We are telling you both. That is the standard we hold the field to. It is also the standard we hold ourselves to.

---

Why this matters

These publications shape CHRO agendas, board decisions, and government policy. When Deloitte says "gradual co-evolution," organisations plan for gradual co-evolution. When Bersin says the talent war continues, organisations compete for talent they should be transitioning. When the WEF says 97 million new jobs are coming, governments defer difficult policy decisions.

The predictions were not wrong because the researchers were careless. They were wrong because the research was built on surveys of organisations in 26 Western economies, optimised for reassurance rather than precision.

The disruption looks different from Dubai than from Davos.

reAImagine.work is not the WEF. It does not have 3,600 executive respondents or 30 books. What it has is a vantage point those publications do not: the view from the regions where the disruption is landing hardest, read through the lens of 30 years of workforce strategy in India, the GCC, and Africa.

Five years from now, someone will write a report card on the predictions being made this week. The question worth asking today is whose data those predictions include.

---

Debu Mishra is the founder of reAImagine.work and Shim Partners Management Consultancy FZCO, based in Dubai. He was a member of the Global Editorial Board for the Deloitte Sensing and Disruption Radar, and the architect of India's Best Boards Research and Awards in partnership with The Economic Times. He can be reached at debu@shimpartners.com