Key Points
- Mass adoption: 69% of companies in major OECD economies already use AI, with the United States leading at 78% — data collected from a panel of 6,000 businesses between November 2025 and January 2026 (Natixis / Atlanta FED).
- LLM dominance: Large Language Models (AI systems trained on vast text datasets) account for 41% of business applications, surpassing classical Machine Learning (statistical pattern-recognition algorithms) stalled at 30%.
- Expanding DTx market: Japan leads the Digital Therapeutics revolution with companies like CureApp, transforming clinically validated algorithms into prescription medical devices on a global scale.
The 69% doesn't lie: AI is already inside companies, and you are falling behind
Enough with the near-future narrative. Artificial intelligence is not coming — it is already here, already working, and already deciding who survives and who disappears. A hard-hitting study commissioned by Natixis Investment Managers and conducted between November 2025 and January 2026 on a sample of 6,000 companies across the four major OECD economies — the United States, the United Kingdom, Germany and Australia — makes this clear. The verdict is blunt: 69% of businesses already use AI. Americans lead at 78%, while Australia closes the group at 59%. These are not projections. These are not optimistic estimates from a pitch deck. These are real numbers, collected in the field.
The detail that should keep you up at night is something else entirely: Large Language Models — LLMs (AI systems generating human-like text from context), for anyone who has been living under a rock — have already dethroned classical Machine Learning, accounting for 41% of business applications against a meager 30% for traditional methods. The transition happened in silence, without fanfare. And inside this transition a trap is hiding: 72% of employees already use AI in the workplace, but 41% of them spend less than one hour per week on it. Office-level experimentation, in other words. Coffee-break curiosity. We are still in the phase where most companies use a Formula 1 tool to do the weekly grocery run.
And here the first structural crack in the system emerges: AI is amplifying the strong and strangling the weak. The companies adopting it most intensively are already the largest, the most productive, the ones paying the highest salaries. Artificial intelligence is not an equalizer — it is an inequality accelerator. For SMEs (small and medium-sized enterprises) that keep stalling, the bill gets heavier with every passing day.
The Singapore paradox: you know everything, you can do nothing
Let us move to Singapore, a global laboratory for HR trends that later become reality everywhere. Local recruiters describe a grotesque situation: recent graduates arrive loaded with theory about artificial intelligence and immediately collide with an invisible wall. Companies are no longer looking for people who can explain what a transformer (a neural network architecture) is — they are looking for people who can implement AI inside real corporate workflows, with all their messiness, their exceptions and their legacy systems dating back to the fax era.
The problem is structural and brutal: junior tasks — data collection, preliminary analysis, drafting documents — have been swallowed whole by AI tools. Young talent is therefore being asked to deliver senior-level operational capability: designing AI architectures, managing complex prompt orchestrations (sequences of instructions guiding AI model behavior), integrating heterogeneous systems. Skills that universities still do not teach, because academic curricula update at the pace of an arthritic tortoise. The result is a generation of graduates who are technically literate but operationally useless for the market that exists today, in 2026. It is not their fault. It is the fault of an educational system that keeps preparing people for a world that no longer exists.
Retail and insurance: AI is not the cute chatbot, it is the demolition of the foundations
There is a colossal misconception circulating in retail company boardrooms: the idea that artificial intelligence means a friendlier chatbot or a virtual fitting room to try on jackets in 3D. Window dressing. Digital cosmetics. The real impact of AI in retail is invisible to the end consumer, because it happens in the back-end (the internal operational layer hidden from users), in the logistical core of the business. Predictive supply chain (AI-driven demand forecasting system) reshapes the supply chain in real time, anticipating demand spikes linked to micro-trends, weather conditions or local events, aiming to reduce warehouse stock to zero. Dynamic pricing (real-time automated price adjustment) no longer simply monitors competitor prices — it processes instantaneous logistical availability and the purchase propensity of specific user clusters. This is industrial surgery, not cosmetic surgery.
In the insurance sector the situation is even more instructive. A concept worth committing to memory has emerged: the "Liquid Claim" (fully automated end-to-end claims processing). The idea is simple in theory, devastating in execution. Having an app that uses Computer Vision (AI that interprets visual data) to recognize a dent on a car body is not enough. If behind that app there are still rows of human authorizations, departmental silos and paper-based processes inherited from the 1990s, AI generates zero ROI. The real revolution is redesigning the entire workflow: automating claim filing, integrating instant anti-fraud checks based on anomalies in historical data, proceeding to automatic settlement for low-risk cases. The result? Processing times collapse from weeks to seconds. Those who do not get there within the next eighteen months will simply be out of the market.
Japan heals with algorithms: welcome to the era of digital therapeutics
The closing chapter of this story comes from the country with the oldest population on the planet, and that is no coincidence. Japan is leading the revolution of Digital Therapeutics — DTx (software-based clinically validated medical treatments) — transforming software into a fully-fledged prescription medical device. The company that symbolizes this movement is CureApp, which does not produce apps for counting steps or reminding you to drink water. It produces clinically validated algorithms that physicians prescribe exactly as they would a drug, to treat hypertension, nicotine addiction and fatty liver disease. The system monitors patient parameters twenty-four hours a day, personalizes cognitive-behavioral interventions (structured therapy targeting thought and behavior patterns) in real time and sends calibrated prompts to correct habits and lifestyle.
The stated objective is not to cure disease — it is to shift the entire healthcare paradigm toward extending healthy years of life, the so-called Healthy Life Expectancy (years lived in good health, not just alive). Fewer hospitalizations, lower systemic costs, more years of productive life. The HealthTech market derived from this is already worth billions of dollars, and projections for the next five years indicate sustained double-digit growth, driven by global demographic aging and the unsustainable pressure on public healthcare systems.
