The book, verifiableEvidence bank
The numbers behind the book
Take nothing on faith. Every key figure in the book is here with its original source and link, so you can verify it yourself.
- Case study95%
2024en vigor (UE 2024/1689)
The EU AI Act enters into force (2024)
On 1 August 2024 the EU AI Act (Regulation 2024/1689) entered into force — the world's first comprehensive legal framework for artificial intelligence. Even so, most of its provisions only apply from 2026: technology runs ahead of the institutions trying to regulate it.
- Data point92%
40% empleo mundial
IMF: 40% of global jobs are exposed to AI
According to the IMF (January 2024), about 40% of global employment is exposed to AI — up to 60% in advanced economies, falling to 40% in emerging markets and 26% in low-income countries. The technology gap between economies also measures the inequality in their capacity to respond.
- Case study92%
2007año (redefine industria)
The iPhone redefines an entire industry (2007)
On 9 January 2007 Apple unveiled the iPhone as a device combining a phone, an iPod and an internet browser. It did not improve the existing phone — it redefined the category and reshuffled the whole industry. This is the exponential-company pattern: it does not compete better, it changes the rules.
- Data point90%
57% aumentar (vs 43%)
Today AI augments more than it automates: 57% vs. 43%
Analysing millions of real Claude conversations, the Anthropic Economic Index found that about 57% of usage is augmentation (collaborating with the person) and 43% automation (fully delegating the task). The nuance matters: “done for you” is, for now, the exception rather than the rule.
- Data point90%
44% competencias a 2027
44% of core skills will be disrupted within five years
In 2023 the WEF estimated that 44% of workers' core skills will be disrupted by 2027 (up from a projected 35% in 2016). The time horizon of change is measured in years, not decades — navigating it demands continuous reskilling.
- Case study90%
14% prod. (+34% novatos)
AI in customer support: +14% productivity, +34% for novices
Across 5,179 support agents, the Brynjolfsson–Li–Raymond study (NBER) found a 14% average productivity gain from introducing a generative-AI assistant, with a 34% improvement for novice workers and almost none for experts. AI spreads best practice and compresses the learning curve.
- Data point90%
100M usuarios (~2 meses)
ChatGPT reaches 100 million users in about 2 months
By a UBS estimate, ChatGPT hit 100 million monthly active users in January 2023, roughly two months after launch — the fastest ramp for a consumer app on record (TikTok took ~9 months; Instagram ~30). It is the canonical illustration of crossing into the exponential moment.
- Case study90%
55,8% más rápido
With GitHub Copilot, coding 55.8% faster
In a controlled GitHub experiment, the Copilot group completed the task (a JavaScript HTTP server) 55.8% faster than the control group: 1h11 vs 2h41. This is human intelligence amplified, not replaced — the person stays in charge, the machine accelerates.
- Case study90%
5,44mil M EUR (a Microsoft)
Nokia sells its phone business to Microsoft for €5.44 billion
The world's leading phone maker until 2007, Nokia failed to answer the iPhone and in September 2013 sold its devices division to Microsoft for €5.44 billion (€3.79bn for devices + €1.65bn for patents; about $7 billion). Its global smartphone share collapsed from ~50% in 2007 to under 5% by 2013. Yesterday's leader can become a cautionary tale within a few years.
- Data point88%
1 → 25 → 59regulaciones de IA · EEUU
U.S. AI regulations went from 1 to 59 in eight years
The number of U.S. federal AI regulations grew from a single one in 2016 to 25 in 2023 and 59 in 2024 (Stanford AI Index). Institutions are chasing a technology society has already adopted — the rule arrives years after the use, exactly the gap this chapter describes.
- Data point88%
109,1 vs 9,3bn USD inversión privada IA
The U.S. invested 12 times more than China in AI in 2024
In 2024 U.S. private AI investment reached $109.1 billion — nearly twelve times China's $9.3 billion and twenty-four times the U.K.'s $4.5 billion (Stanford AI Index). The technology gap is not only about capabilities but about capital: it defines who can respond to change and who merely endures it.
- Data point88%
170 / 92M empleos (neto +78M)
By 2030: 170 million jobs created, 92 million displaced
The WEF Future of Jobs 2025 projects 170 million new jobs created and 92 million displaced by 2030 — a net +78 million — with 39% of core skills set to change. Not the end of work but its redefinition: creating, working, learning and governing under new rules.
- Data point88%
25% prima salarial IA
AI skills command up to a 25% wage premium
PwC's 2024 Global AI Jobs Barometer, analysing over 500 million job ads across 15 countries, found that roles requiring AI skills carry a wage premium of up to 25% (US average), and that productivity grew nearly five times faster (4.8x) in the sectors most exposed to AI. The skills race already has a market price.
- Case study88%
2023año (prohibición interna)
Samsung bans ChatGPT for staff after a data leak
In May 2023 Samsung banned employees from using generative-AI tools after discovering sensitive internal source code had been uploaded to ChatGPT. The trust challenge is concrete: what you share with an external model can fall out of your control.
- Data point88%
6 de cada 10necesitan formación 2027
6 in 10 workers will need training before 2027
The WEF estimates six in ten workers will require additional training before 2027, yet only half currently have access to adequate training opportunities. Anticipating — training before the need turns urgent — is the difference between getting ahead and merely reacting.
- Data point88%
30 → 9 → 2meses a 100M usuarios
Time-to-100-million-users keeps collapsing: 30 → 9 → 2 months
Instagram took about 30 months to reach 100 million users; TikTok about 9; ChatGPT just 2 (UBS). The same milestone is reached in ever less time — the pattern the three-T paradigm sets out to capture: acceleration is not linear, it accelerates on itself.
- Data point85%
nº 1 de 25factores de impacto en EBIT
Redesigning workflows: the #1 factor for AI to move EBIT
Of 25 attributes tested, McKinsey finds that redesigning workflows correlates most with real AI impact on EBIT; 'high performers' are nearly three times more likely to have fundamentally redesigned them. Bolting AI onto the old way isn't enough — you have to operate the system, which is exactly what this manual lays out.
- Data point85%
1% de empresas «maduras» en IA
Only 1% of companies consider themselves 'mature' in AI
Almost every company invests in AI and 92% plan to increase spending over the next three years, yet only 1% of leaders rate their deployment as 'mature' — AI woven into workflows and driving real outcomes (McKinsey). The bottleneck isn't the technology or the workforce but the lack of direction: precisely why you need a compass before the journey.
- Data point85%
> 25% del código nuevo (Google)
At Google, AI already generates over 25% of new code
On the Q3 2024 earnings call, Sundar Pichai said “today, more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers.” Creating and working stop being what they were: the first of the four redefinitions is no longer theory — it is happening inside the world's largest software company.
- Case study85%
25,6M USD (fraude por deepfake)
A deepfake video call cost $25 million
In 2024, an employee of engineering firm Arup in Hong Kong made 15 transfers totalling $25.6 million after a video call in which the “CFO” and every colleague were deepfake recreations (CNN). When seeing and hearing stop being proof of truth, trust — and its ethical governance — becomes the critical infrastructure.
- Data point85%
75 / 78% usan IA / traen la suya
75% already use AI at work; 78% bring their own
75% of knowledge workers already use AI at work — usage nearly doubled in six months — and 78% bring their own tools (BYOAI), often without their company knowing (Microsoft & LinkedIn, 2024). Exponential people adopt AI bottom-up, ahead of their organisations.
- Case study85%
100M usuarios mensuales
Duolingo passes 100 million monthly learners
Duolingo ended 2023 with 88.4 million monthly active users and crossed 100 million soon after, democratising language learning through gamification and, later, AI. An example of the exponential person at scale: tools that once needed an academy now fit on anyone's phone.
- Case study85%
+427% interanual · centro datos
NVIDIA anticipated the compute wave: +427% in one year
NVIDIA bet on accelerated computing for AI years before demand exploded. When it arrived, its fiscal-Q4-2024 data-center revenue hit $22.6 billion, up 427% year over year. Getting ahead — not reacting — is what turned a technical bet into the defining rent of the AI era.
- Data point85%
2,6–4,4billones USD/año
Generative AI could add $2.6–$4.4 trillion per year
Across 63 use cases, McKinsey estimates generative AI could add between $2.6 and $4.4 trillion in annual value to the global economy. It puts a monetary scale on the forces of change this chapter describes — not an abstract promise but a measurable economic magnitude.
- Data point85%
300M empleos · +7% PIB
Generative AI could expose 300 million jobs and raise GDP by 7%
Goldman Sachs estimated generative AI could automate the equivalent of 300 million full-time jobs worldwide while raising global GDP by around 7% over a decade. Two faces of the same force: displacement and growth.
- Case study85%
2012año de la quiebra
Kodak invented the digital camera in 1975 and went bankrupt in 2012
Kodak engineer Steven Sasson built the first digital camera in 1975, but the company protected its film business rather than cannibalise it. In January 2012 it filed for Chapter 11 bankruptcy. The adaptation crisis is not failing to see change — it is seeing it and not daring to embrace it.
- Case study85%
2010año de la quiebra
Blockbuster files for bankruptcy in 2010, beaten by Netflix
Blockbuster operated around 9,000 stores and in 2000 turned down buying Netflix for $50 million. In September 2010 it filed for Chapter 11 bankruptcy: change was already happening for those who refused to see it. The textbook case of business-model disruption.
- Case study82%
3.300 → 11.000+licencias ChatGPT (BBVA)
BBVA scaled from 3,300 to over 11,000 generative-AI licences
In May 2024 BBVA — the first European bank to partner with OpenAI — rolled out 3,300 ChatGPT Enterprise licences and, on internal demand, expanded to 11,000, on its way to the whole workforce across 25 countries; over 80% of users use it daily. An exponential enterprise doesn't pilot AI in a corner — it operates it at scale.
- Data point80%
70% de habilidades para 2030
70% of a job's skills will change by 2030
The skills required for the same role have already shifted by about 25% since 2015, and LinkedIn projects that by 2030, 70% of the skills used in most jobs will have changed, with AI as the catalyst. The skills race isn't won once: resilience is learning to relearn.
- Case study80%
99% servicios públicos online
Estonia delivers 99% of its public services online
Estonia provides 99% of its public services digitally on top of digital ID and the X-Road infrastructure; with e-divorce (Dec 2024) it reached effectively 100%. The government estimates digital signatures save the equivalent of 2% of GDP — roughly 800 years of working time — per year. An exponential society reorganises the state, not merely digitises paperwork.
- Data point78%
70+países con políticas de IA
Over 70 countries now have national AI policies
The OECD.AI observatory tracks AI policy initiatives from more than 70 countries and jurisdictions, and around 41 already have a national strategy aligned with the OECD principles. Exponential societies don't wait: they organise AI governance while the technology is still rolling out.