By 2036, AI may feel less like a separate tool and more like a layer built into everyday software, services and physical devices. The biggest change may not be one dramatic invention, but millions of ordinary decisions supported—or sometimes made—by machines.
That future is not predetermined. Technical breakthroughs, energy availability, regulation, public trust, business choices and social priorities will all influence what develops. The following outlook describes possibilities, not promises.
1. AI assistants could become persistent collaborators
Today's assistants usually wait for a prompt. Over the next decade, they may retain approved context, coordinate across tools and follow longer projects. A personal system could help plan learning, organize information, prepare communications and monitor goals while adapting to a user's preferences.
The value will depend on control. Useful memory requires careful consent, visible data boundaries and simple ways to inspect or delete stored information. People should be able to decide which actions require confirmation and which can happen automatically.
2. Work may be redesigned around human–AI teams
AI is likely to change tasks faster than entire occupations disappear. Routine drafting, analysis, scheduling and documentation may become increasingly automated. Many jobs could shift toward defining problems, supervising workflows, checking quality and handling situations that require trust, negotiation or domain judgment.
This transition will not affect everyone equally. Organizations that invest in training and redesign work thoughtfully may expand employee capability. Those that treat AI only as a cost-cutting mechanism risk fragile processes, lost expertise and poor accountability.
Work
AI becomes a standard collaborator across knowledge work, with people directing goals and reviewing consequential outputs.
Education
Tutoring and practice become more personalized, while teachers remain essential for motivation, context and student wellbeing.
Science
Models help researchers search complex possibilities, design experiments and connect discoveries across disciplines.
Healthcare
AI supports administration, detection and treatment planning under professional oversight and strong privacy controls.
3. Education could become more adaptive
AI tutors may offer explanations at different levels, generate practice around a learner's mistakes and make high-quality support available beyond classroom hours. Teachers could spend less time producing routine materials and more time guiding discussion, motivation and deeper understanding.
Assessment will need to evolve. When generating a polished answer is easy, schools may place more emphasis on live explanation, projects, process evidence and the ability to evaluate AI output. AI literacy—including when not to use AI—could become as fundamental as digital literacy.
4. Scientific discovery may accelerate
AI can help researchers examine enormous search spaces, identify patterns across papers and propose candidates for further testing. Over ten years, these systems may become stronger partners in materials research, biology, medicine, climate modelling and engineering.
AI will not remove the need for experiments or scientific skepticism. Generated hypotheses still require evidence. The most valuable systems may be those that connect computation with laboratories while documenting uncertainty and making results easier to reproduce.
5. Robots may leave controlled environments
Improvements in perception, language-based instruction and planning could make robots more adaptable. Warehouses and factories will remain important, but capable systems may expand into agriculture, inspection, logistics, construction and selected forms of care.
Physical reality is unforgiving. A chatbot can revise a sentence; a robot's mistake can damage property or injure someone. Progress outside controlled environments will depend on hardware reliability, affordable energy, rigorous testing and safe fallback behavior.
6. Synthetic media may require visible proof
Generated video, audio and images are likely to become more realistic and easier to produce. Creative expression will expand, but distinguishing authentic records from manipulation will become harder. Provenance systems, disclosure labels and verification habits may become a routine part of digital communication.
Regulation is already moving in this direction. The European Union's AI framework includes transparency requirements for certain generated or altered content, showing how disclosure could become a normal expectation rather than an optional courtesy.
7. Governance may become part of product design
The next decade will likely bring a patchwork of laws, technical standards and industry rules. High-impact uses in healthcare, employment, education, finance and public services will face greater demands for testing, documentation, human oversight and appeal.
Good governance does not have to prevent innovation. Clear risk levels can allow low-risk uses to move quickly while requiring stronger evidence where systems can affect rights or safety. Frameworks such as NIST's AI Risk Management Framework already encourage organizations to address trustworthiness throughout an AI system's lifecycle.
8. Energy and computing will become strategic constraints
More capable models and wider AI adoption require data centres, chips, grids and cooling. The International Energy Agency projects electricity generation for data centres to rise from 460 TWh in 2024 to more than 1,000 TWh in 2030 in its base case. Efficiency gains, renewable generation and careful choices about where AI adds real value will matter.
This pressure could also drive innovation: smaller models, specialized chips, local processing and systems that use computation only when needed. The best future AI may not always be the largest; it may be the system that achieves the right outcome with the least cost and energy.
Intelligence has a physical footprint
AI may feel digital, but it depends on physical infrastructure. Energy, water, minerals, manufacturing and network capacity will shape which systems are affordable and sustainable.
9. AI could become more capable without becoming AGI
Systems may improve greatly at reasoning, memory, tool use and multimodal interaction while still remaining unreliable in unfamiliar situations. A decade of progress does not guarantee Artificial General Intelligence, and AGI would not automatically mean conscious or superintelligent AI.
Forecasts should therefore separate observable capability from labels. Useful questions include: Can the system complete new tasks reliably? Does it recognize uncertainty? Can people understand and control its actions? How does it behave when conditions change?
10. Human skills may become more valuable, not less
As machines handle more production, people may contribute most through direction, judgment and relationships. The ability to frame a worthwhile problem, combine knowledge from different fields, communicate with others and take responsibility for a decision will remain difficult to automate completely.
Continuous learning will matter because tools and workflows will keep changing. The goal is not to memorize every new product. It is to develop durable foundations: subject knowledge, critical thinking, data awareness, ethical judgment and the confidence to learn unfamiliar systems.
Three scenarios for 2036
Measured progress
AI becomes more useful and efficient, but reliability limits keep human oversight central in complex work.
Broad acceleration
Agentic systems transform many workflows, scientific discovery advances and economic adaptation becomes a major priority.
Uneven transformation
Some regions and industries advance rapidly while infrastructure, skills, policy and access create wide differences elsewhere.
The decade ahead
The next ten years of AI will be shaped as much by human choices as by technical capability. The most successful future will not simply contain smarter systems—it will use them to make people more capable, informed and empowered.