What the World’s Most Powerful Business Leaders Really Think About AI

When a technology revolution becomes a personal question

AI can write, calculate, design, research, translate, analyse and increasingly act on instructions. This raises a question that is no longer limited to technology conferences: what happens to us when intelligence itself becomes widely available?

The people building the world’s largest AI companies offer very different answers. Some see abundance and faster discovery. Some see a new industrial age. Others openly warn about job losses and economic disruption. Taken together, their statements tell a more interesting story than either ‘AI will save the world’ or ‘AI will take every job.’

When intelligence becomes widely available, the question may no longer be who can produce the most work—but who knows what work is worth doing.

Jensen Huang: job security will come from using AI

NVIDIA CEO Jensen Huang has become one of the most quoted voices of the AI age. His broad message is simple: AI may not directly take your job, but a person who knows how to use it may have an advantage over someone who does not.

Huang points to radiology, where AI can examine scans rapidly. This has not removed the need for radiologists. It can help them handle more cases and spend more time with patients. He makes a similar point about nurses: if AI handles charting and transcription, nurses can spend a larger share of their day caring for people.

A job is rarely one task. A teacher does not merely prepare notes. An accountant does not merely enter numbers. A marketer does not merely write captions. Most jobs mix tasks with judgement, communication, responsibility and human contact. AI may take over parts of a role without removing the complete role.

The better career question is not simply, ‘Will AI replace my profession?’ It is: which parts of my work can AI perform, and what should I become better at once those parts are automated?

Satya Nadella: natural language is becoming the new computer interface

Microsoft CEO Satya Nadella describes AI as a change in how people interact with computers. Microsoft’s Copilot strategy rests on the belief that natural language may become a universal interface.

Instead of learning where every button is located or memorising every formula, people can describe what they want: compare this month’s sales with last month; summarise the risks in this contract; turn meeting notes into an action plan; find the reason behind a fall in conversions.

Earlier computer revolutions rewarded people who learned the language of machines: commands, menus, formulas or code. The AI revolution is asking machines to learn the language of people.

But conversational ease can create false confidence. A polished answer may still be wrong. A spreadsheet formula may look convincing while referring to the wrong cells. Natural language reduces the difficulty of operating software; it does not remove the need to understand the work.

Sundar Pichai: AI will spread far beyond the technology industry

Google and Alphabet CEO Sundar Pichai sees AI as a general-purpose technology whose effects will reach science, education, healthcare, government and everyday business. In a 2025 speech, he described a coming ‘golden age of innovation’ while arguing that its benefits will depend on infrastructure, skills, policy and access.

A calculator is useful in certain tasks and a camera in others. AI can sit inside almost every form of software and take part in thousands of activities. A retailer may use it to study demand. A hospital may use it to reduce documentation. A manufacturer may use it to predict equipment failure. A student may use it as a tutor.

The future may therefore look less like one dramatic robot entering the office and more like intelligence quietly appearing inside every tool we already use.

Andy Jassy: some jobs will shrink while new ones appear

Amazon CEO Andy Jassy has given one of the more direct corporate assessments. As Amazon introduces generative AI and agents, he expects the company to need fewer employees for some forms of work and more employees for others. He has also said that AI-related efficiency could reduce Amazon’s total corporate workforce over the next few years.

This is less comforting than the common promise that AI will only assist workers. Both outcomes can happen at once: AI can make one employee far more productive and still cause a company to hire fewer people for that function.

History shows that technology can create more work over time while causing pain during the transition. New jobs do not always reach the same people, in the same cities, at the same salaries or at the right time.

• Which entry-level roles will remain?
• How will beginners gain experience if AI performs much of the junior work?
• Who will pay for reskilling?
• Will productivity gains be shared with workers?
• How quickly can education respond?

Dario Amodei: imagine a country of geniuses inside a data centre

Anthropic CEO Dario Amodei uses a striking image for highly capable AI: a ‘country of geniuses in a datacenter.’ He warns that such systems could bring the largest change to the global labour market in human history.

Today, many people think of AI as a chatbot that helps write an email. Amodei asks us to imagine millions of capable digital workers operating at computer speed. What happens when a company can call upon near-instant researchers, analysts, programmers and planners? What happens when the supply of certain forms of intelligence is no longer limited by the number of trained people available?

The price of many services could fall. Scientific discovery could speed up. Small companies could perform work that once required large teams. At the same time, power may become concentrated among organisations controlling the models, computing capacity, energy and data.

The future of AI is therefore not only a story about clever software. It is also a story about ownership.

Marc Benioff: managers may soon lead humans and AI agents together

Salesforce CEO Marc Benioff has described agentic AI as a new labour and economic model. He has also said that today’s leaders may be the last generation to lead a human-only workforce.

An AI agent is different from a chatbot that waits for one question. An agent may receive a goal, break it into steps, use connected software, retrieve information, prepare an output and request human approval when needed.

A sales manager may supervise people alongside agents that research prospects, update records and prepare follow-ups. A finance manager may use agents to collect data, flag unusual transactions and draft variance notes. A training institute may use agents to classify enquiries, prepare counsellor notes and schedule follow-ups.

The manager’s job then shifts from assigning every small task to designing a working system: what may the agent do independently, what data may it access, when must it stop, who checks its work and who remains responsible when it makes a mistake?

Sam Altman: human judgement matters more as AI grows stronger

OpenAI’s public plan for advanced AI states that faster technical progress makes human judgement and public coordination more important, not less. It argues that the future should be shaped by people and societies rather than only by companies building the strongest systems.

AI companies compete on model ability, computing power and agents. Businesses compete to reduce cost and move faster. Countries compete to build infrastructure and protect their place in the global economy. Yet the fastest available option is not always the wisest one.

A system that can make decisions still needs limits. A system that can draft medical, financial or legal recommendations still needs accountable human review. A system that can observe workers and customers raises questions about privacy and consent.

The more capable AI becomes, the more carefully we must decide where it should not act alone.

What does this mean for an ordinary professional?

These CEOs disagree about timelines and consequences, but their statements point in the same direction. AI use will become an ordinary workplace skill rather than something limited to programmers. Subject knowledge will still matter because only someone who understands the work can judge whether an answer is fit for use.

Repetitive digital tasks face the most pressure. People whose role depends mainly on routine preparation may need to move towards interpretation, customer contact, decision-making, ownership or specialist knowledge.

Learning one AI tool will not be enough because products will keep changing. The lasting ability will be knowing how to define a problem, give context, inspect the result and improve the process.

Human qualities may become more important precisely because machine intelligence is becoming cheaper. Nadella once argued that in a world filled with artificial intelligence, human qualities such as empathy would become scarce.

When machines can produce more words, trust matters more. When machines can generate more options, judgement matters more. When machines can perform more tasks, purpose matters more.

The AI5 Sunday thought

AI probably will not arrive one morning and suddenly replace the working world. It will enter task by task, tool by tool and decision by decision.

Job security will not come from pretending this change is temporary. Nor will it come from accepting every AI answer without question.

It will come from learning how to work with AI while becoming better at the parts of work that require context, responsibility, judgement and humanity.