The Future of AI: What's Coming in 2026 and Beyond

The scale of what's happening

$200B

Projected AI infrastructure investment in 2025 alone — from US hyperscalers, Stargate consortium, and Chinese state investment [Goldman Sachs]

2027

When OpenAI, Anthropic, and several major labs say they expect to reach or approach AGI — with significant uncertainty on both sides [Lab statements]

85%

Of jobs will be significantly impacted by AI by 2030 according to McKinsey — but "impacted" includes being augmented, not just automated [McKinsey Global Institute]

The trends that will define the next 5 years

2025–26

Agentic AI becomes mainstream

AI systems that don't just answer questions but execute multi-step tasks autonomously — booking appointments, writing and running code, managing files, browsing the web, calling APIs. Operators (OpenAI), Claude Computer Use (Anthropic), and Gemini's Android integration are early versions. Within two years, the default interaction with AI for work tasks will be agentic: you describe an outcome, the AI executes the steps. Human-in-the-loop checkpoints will replace human execution for most knowledge work tasks.

2025–27

Multimodal becomes the norm

The distinction between text AI, image AI, audio AI, and video AI disappears. Models process and generate across all modalities natively. GPT-4o's voice mode was the preview — you speak, it responds naturally with emotion and context. Next generation: continuous video understanding, real-time translation across media, AI that understands your environment through a camera and responds verbally. The AI interface becomes ambient rather than screen-bound.

2026–28

Physical AI: robots and embodied intelligence

Figure AI, Boston Dynamics, Tesla Optimus, and 1X Technologies are all deploying humanoid robots with AI brains trained on the same techniques as LLMs. The breakthrough is generalisation — previous industrial robots followed fixed programs, new ones learn from demonstration. By 2028, humanoid robots will be in commercial deployment for warehouse and manufacturing tasks at scale. This extends AI's economic impact from digital into physical labour markets.

2026–30

Scientific AI accelerates discovery

AlphaFold solved protein structure prediction. AlphaGeometry matched IMO gold medalists. AlphaProof proved mathematical theorems. The pattern: AI rapidly reaching expert-level capability in specific scientific domains where problems are well-specified and success can be objectively measured. Drug discovery (Insilico Medicine, Recursion Pharmaceuticals), materials science, and climate modelling are next. The AI-accelerated science era means the pace of scientific progress will increase substantially — with hard-to-predict downstream effects.

2026–29

AI economics reshape labour markets

The productivity gains are real and unevenly distributed. Workers who use AI well become more productive and valuable. Workers who don't, in roles that AI can replace, face displacement pressure. The historical precedent from previous automation waves suggests new job categories emerge — but the pace of AI advancement is faster than previous technological transitions, which reduces the time for labour market adaptation. The IMF's 2024 analysis found AI likely to affect 40% of jobs globally, with advanced economies more exposed than emerging ones due to higher knowledge work concentration.

2026–30

Regulation and governance catch up

The EU AI Act is in full enforcement from 2026. The US is developing sector-specific AI regulations after years of voluntary commitments from labs. China has AI content and algorithm regulations already active. The international coordination challenge — preventing regulatory arbitrage where AI development moves to less-regulated jurisdictions — will define AI policy for the decade. The decisions made about AI governance in the next 3-5 years will shape the technology's development trajectory for a generation.

Three scenarios for 2030

Optimistic

The productivity supercycle

AI delivers the largest productivity increase since the industrial revolution. Cancer detection improves dramatically. Drug discovery timelines collapse. Climate solutions accelerate. AI tutors democratise education access. The economic gains are distributed broadly enough to offset displacement, and new job categories emerge faster than old ones disappear. AI alignment progress keeps pace with capability improvements.

Most likely

Uneven transformation

Significant productivity gains in certain sectors. Meaningful job displacement concentrated in specific roles (white-collar knowledge work, entry-level tasks). Widening skills and income gap between AI-leveraged workers and those who aren't. Genuine scientific breakthroughs. Ongoing AI safety concerns that are manageable but not resolved. A world that looks dramatically different from 2024, but not utopian or dystopian.

Risk scenario

Concentrated power and disruption

AI capabilities significantly outpace alignment and governance. Rapid job displacement without adequate social support systems creates political instability. AI capabilities concentrated in a small number of companies and states, increasing inequality between nations. Misuse of AI for surveillance, manipulation, and autonomous weapons becomes widespread. The "moving fast" ethos produces preventable harms that erode public trust.

What you should actually do about this

The future of AI isn't something that happens to you — it's something you participate in shaping through how you use, evaluate, and engage with it. At the individual level: build AI literacy now, before your field requires it under pressure. Understand what AI can and can't do, not just how to use it. At the professional level: identify the 20% of your work that AI can augment and start there — don't wait for your employer to mandate it. At the societal level: the governance decisions happening now — about how AI systems are deployed, what data they use, who has access, and who's accountable for harms — matter more than the technical benchmarks. The people who engage critically and substantively with those decisions will have more influence over how AI develops than those who watch from the sidelines.

FAQ

When will AI reach AGI (Artificial General Intelligence)?

The honest answer is: it depends how you define AGI, and no one knows. OpenAI CEO Sam Altman has said he believes AGI "in the traditional sense" could arrive within a few years. Anthropic's Dario Amodei has expressed similar views. Sceptics argue current LLM-based approaches have fundamental limits that preclude true general intelligence. The disagreement partly reflects definitional differences — if AGI means "can do any cognitive task a human can do," that's a higher bar than "is economically significant enough to transform most industries." The second bar — economic transformation — is arguably already being crossed.

Should I be worried about AI taking my job?

Depends on your role. Roles most at risk: entry-level data analysis, routine writing tasks (standardised reports, copy), certain legal and accounting tasks (document review, basic compliance), and customer service triage. Roles with lower near-term risk: roles requiring physical presence, deep human relationships, creative judgment, novel problem-solving, and complex domain expertise. The most durable position: become the person who uses AI tools better than anyone else in your field. Every professional whose work AI touches has a choice between being displaced by AI and being amplified by it — and being early to build those skills provides a meaningful advantage while the adaptation curve is still steep.

The Future of AI: What's Coming in 2026 and Beyond

04 — Don't watch from the outside