REPORT: The State of Artificial Intelligence Adoption

Nov 7

Lessons from the Past, Signals from the Present, and What Comes Next

Every major technological shift feels unprecedented when you are living through it. Yet, once the dust settles, the same patterns always emerge; fear, over-investment, correction, and then integration.

Artificial Intelligence (AI) now stands at that midpoint: the hype is cooling, the correction is underway, and integration is beginning. To see where this goes next, we must look backward at earlier revolutions, understand the roots of AI itself, and examine what its spread means for people, businesses, and industries worldwide.

Defining Artificial Intelligence

Artificial Intelligence (AI) refers to computer systems designed to perform tasks that typically require human intelligence, understanding language, recognising patterns, solving problems, and learning from data.

Early AI was symbolic, built from human-coded rules. It evolved into machine learning (ML), where algorithms improved by analysing data. In 2017, a key shift arrived with the transformer architecture, allowing large systems to process context and meaning at scale. This paved the way for Large Language Models (LLMs) and Generative AI, which can now write, code, analyse, and create with remarkable fluency.

How History Repeats Itself

The Looms and the Luddites

In the early 1800s, textile workers known as Luddites smashed mechanised looms that threatened their livelihoods. They were not anti-technology; they were protesting sudden change. Within a generation, the weaving industry transformed, productivity soared, and new types of work appeared. Resistance, it turns out, is part of progress.

Electricity and the Productivity J-Curve

When factories switched from steam to electric power, output hardly changed at first. Only after managers redesigned entire workflows around decentralised motors did productivity surge. Economists call this the J-curve effect: a dip before the lift. True adoption always requires reorganisation, not just new tools.

ATMs, CAD, and RPA

Each wave of automation follows a similar pattern.

Automated Teller Machines (ATMs) changed banking, but tellers did not disappear, they shifted to sales and customer service.
Computer-Aided Design (CAD) software improved engineering output, but only after training and process redesign.
Robotic Process Automation (RPA) promised instant efficiency, then created governance issues when poorly managed.

The Internet and the Dot-Com Era

The 1990s internet boom ended in a crash. Yet once infrastructure, standards, and business culture matured, digital technology quietly became essential. The same trajectory is visible today with AI.

From Early AI to the Generative Era

The 1950s–1980s: Expert Systems

The first wave of AI relied on rule-based expert systems that captured specialist knowledge in code. Key examples include:

DENDRAL (1965), which interpreted chemical structures for researchers.
MYCIN (1972), which diagnosed bacterial infections and recommended treatments.
XCON (1980), used by Digital Equipment Corporation to configure computer systems.

These systems worked well in narrow domains but failed to generalise. When maintenance costs rose and computing power plateaued, the first “AI winter” arrived.

IBM Watson and the Quest for Commercial AI

In 2011, IBM Watson captured public imagination by defeating human champions on Jeopardy!. Its commercial expansion into healthcare revealed both promise and limits: deep data insights, but costly integration. The lesson was clear, intelligence without implementation does not equal impact.

The Transformer Revolution and OpenAI’s Spark

In 2017, Google researchers introduced the transformer model, a breakthrough architecture that powered modern generative AI. Within a few years, OpenAI’s Generative Pre-trained Transformer (GPT) models demonstrated natural language fluency and accessibility.

As OpenAI’s research culture spread, new companies emerged, Anthropic, Cohere, Mistral, and others, building on open knowledge and shared infrastructure. By 2025, the AI ecosystem had become decentralised, with open-source models and national AI strategies reshaping the field.

The Current State of AI Adoption

Surveys through 2024–2025 show roughly two-thirds of large enterprises now use AI or Generative AI in some form. North America leads in adoption, Europe leads in regulation, and developing markets are experimenting fast.

The industry is in its correction phase:

Rapid startup expansion is giving way to consolidation.
Enterprise pilots are turning into operational systems.
Decision-makers now ask how to measure return on investment (ROI), energy use, and compliance risk.

Running a single AI query still costs several times more than a basic web search, but hardware efficiency and model optimisation are improving rapidly. As with electrification and cloud computing, costs will drop as standardisation spreads.

The Seven Friction Points of Adoption

Cost — Compute and energy expenses remain high. AI is a capital investment, not a quick cost-saver.
Skills — Roles evolve faster than people. Upskilling is more effective than replacement.
Governance — Unchecked pilots lead to “shadow AI.” Guardrails create trust and scale.
Culture — Technology adoption is human first. Leadership alignment matters as much as tools.
Productivity Lag — Every revolution dips before it lifts. Patience through the J-curve delivers results.
Regulation — Frameworks are tightening globally. Early compliance builds long-term trust.
Fear — Every disruption begins with anxiety. Organisations that learn together move forward faster.

These friction points are not obstacles but waypoints in a normal adoption curve.

Turning Friction Into Progress

Each pain point carries its own solution:

Cost drives efficiency. Smarter, smaller models reduce waste.
Skills gaps spark new professions, prompt engineering, AI auditing, data stewardship.
Governance forces process redesign, which builds resilience.
Cultural pushback encourages transparency and participation.

The correction phase is not decline. It is calibration.

Humans and AI Working Together

AI adoption works best when it extends human ability instead of replacing it. Real-world examples show this balance clearly:

Design — AI drafts layouts; people refine ideas and emotion.
Logistics — Algorithms plan routes; drivers adapt in real time.
Customer Service — Chatbots handle simple queries; humans solve complex issues.
Finance — Models flag anomalies; analysts investigate and act.
Healthcare — AI analyses scans; doctors diagnose and treat patients.
Education — AI personalises learning; teachers connect with students.

AI scales repetition. Humans provide judgment. Together they perform better than either alone.

Regional Perspectives

United States

High corporate adoption and venture capital investment. Regulation remains light but is shifting toward transparency and accountability. Ethical and environmental debates dominate public discourse.

United Kingdom

The Bletchley Declaration (2023) positioned the UK as a mediator between innovation and safety. The country emphasises flexible governance and international collaboration.

European Union

The AI Act (effective 2024) sets the world’s first comprehensive AI regulatory framework. Implementation between 2024 and 2027 will establish global norms for risk management and transparency.

South Africa and Africa

South Africa’s National Artificial Intelligence Policy Framework (2024) focuses on responsible innovation and skills development. The African Union’s Continental AI Strategy aims for inclusive growth by 2030.

China

China treats AI as strategic infrastructure. National plans emphasise sovereign data, large-scale deployment, and leadership in AI hardware. State investment is driving rapid implementation across manufacturing, logistics, and governance.

Asia-Pacific Region

Across Asia, nations such as Singapore, Japan, and South Korea are investing heavily in AI ethics, education, and research. Singapore’s Model AI Governance Framework remains a global benchmark for practical policy.

The Hybrid Future

AI will soon be invisible, part of every workflow, every interface, every decision. The next stage is hybrid intelligence, where people and machines collaborate by design.

Short-term (1–3 years): consolidation, cost management, and clear governance.
Long-term (5–10 years): seamless human-AI teams where technology supports creativity, insight, and ethical decision-making.

Sustainability will become central. Organisations will track energy per model and emissions per query alongside ROI.

Implications for Traditional Industries

Every past revolution divided incumbents into two groups: those who adapted workflows and those who did not.
Factories that re-engineered for electricity survived; those that clung to steam vanished.
Retailers that embraced barcodes gained efficiency; those that resisted fell behind.

For today’s banks, hospitals, logistics firms, and governments, the lesson is identical: treat AI not as a gadget but as infrastructure. Build the systems, data, and culture to support it, or risk being outpaced by those who do.

Conclusion

Artificial Intelligence adoption follows a pattern older than itself. We invent, we resist, we correct, and we integrate. The turbulence of 2025 marks not the end but the midpoint of the J-curve, the messy, necessary process before stability and growth.

The companies that stay the course, invest in people, and design for hybrid collaboration will define the next decade. AI will not replace work; it will reshape it.

Innovation always feels chaotic up close. With perspective, it looks inevitable.

Call to Action: Partnering With Mill Collective

At Mill Collective, we help organisations navigate the human and strategic side of technology change. From research and readiness assessments to AI adoption roadmaps and content strategy, our work bridges insight and implementation.

If your business is ready to turn AI friction into forward motion, connect with us.
Together, we can design the workflows, skills, and stories that make technology work for people — not the other way around.

Visit [Mill Collective’s site] or reach out directly to start your next chapter of intelligent transformation.

Sources

McKinsey & Company, “The State of AI: Global survey,” 2024. McKinsey & Company+2McKinsey & Company+2
KPMG International, Generative AI Adoption Index Report, November 2024. KPMG Assets
Deloitte AI Institute, The State of Generative AI in the Enterprise 2024. Deloitte Italia
Boston Consulting Group (BCG), “AI Adoption in 2024: 74 % of Companies Struggle to Achieve and Scale Value,” October 2024. Boston Consulting Group
PwC, 2024 US Responsible AI Survey. PwC
Stanford University, 2024 AI Index Report. Stanford HAI
World Economic Forum & Accenture, China’s Path to AI-Powered Industry Transformation, January 2025. World Economic Forum Reports
Georgetown University Center for Security and Emerging Technology, “An Analysis of China’s AI Governance Proposals.” CSET
Organisation for Economic Co‑operation and Development (OECD).AI, “AI Strategies and Policies in China.” OECD AI
RAND Corporation, “China’s Evolving Industrial Policy for AI.” RAND Corporation
Wharton School, University of Pennsylvania & GBK Collective, “Growing Up: Navigating Generative AI’s Early Years,” October 2024. Wharton Human-AI Research
Mayer Brown, “Artificial Intelligence: A Brave New World – China Formulates New AI Global Governance Action Plan,” October 2025. Mayer Brown
China’s Ministry of Foreign Affairs, “AI Capacity-Building Action Plan for Good and for All,” September 2024. Chinese Foreign Ministry
Tech Society blog, “China’s AI Strategy and Insights Into 2025,” September 2024. digitalstrategy-ai.com
Riverbed Technology, “Global AI & Digital Experience Survey 2024,” 2024. Riverbed

Simon Moolman