AI Predictions: The Next 1, 2, 5 and 10 Years
Our view of where artificial intelligence is heading, told as one story across four time horizons, with twenty detailed predictions behind each one.
The short version: in the next year AI becomes dependable enough to finish real tasks on its own. Within two years it takes on whole projects. Within five, organisations are rebuilt around it and some office jobs clearly shrink. Within ten, most work done on a computer is done by AI, and what people are paid for shifts towards judgement, trust and care.
These are our own forecasts, written in November 2026. They are informed opinion, not fact, and some will be wrong. Each prediction carries a confidence rating of High, Medium or Low, and confidence falls the further ahead we look. Nothing here is career, financial or investment advice.
AI that finishes the job
Over the next twelve months the biggest change is reliability, not new tricks. Agents that need watching today will be trusted with tasks that run for hours, such as research, coding and tidying data, and will hand back work they have already checked. Speaking to an assistant becomes as normal as typing, and more of it happens privately on your own phone or laptop.
At work, the agent moves inside the software people already use, so you hand over a task instead of hunting for a feature. Customer service and software development change first. For jobs, the early sign is quieter than headlines suggest: fewer junior openings in routine office roles, and new work for the people who set up and supervise agents.
- AI agents reliably finish multi-hour tasksAI capabilities · High confidence
- Most office software has an agent built inWork · High confidence
- Assistants remember youEveryday users · High confidence
- Entry-level office hiring tightensJobs · Medium confidence
From single tasks to whole projects
Two years out, AI stops being something you ask for one answer at a time. Agents plan and carry out projects that last days, checking in at decision points, and they can operate ordinary software the way a person does. One employee overseeing several agents becomes a recognisable way of working.
That changes how work is priced and organised. Clients question hourly billing, very small companies take on work that used to need a department, and regulated industries adopt AI with the audit trails that new rules demand. Roles in support and administration are merged, while people who pair real expertise with skilled use of AI are paid noticeably more.
- Agents handle multi-day projectsAI capabilities · Medium confidence
- One person manages a team of agentsWork · Medium confidence
- An assistant acts across your accountsEveryday users · Medium confidence
- Support and admin roles are redesignedJobs · Medium confidence
Organisations rebuilt around AI
By 2031 we expect AI to match experts at most tasks done entirely on a computer, and to be far cheaper than it is now. The harder part, whole jobs with judgement, context and accountability, remains human for longer. Robots leave the laboratory for warehouses, factories and some hospitals.
Organisations change shape: fewer layers, smaller teams, and agents treated like staff accounts with their own permissions and logs. This is where displacement in clerical and support work becomes measurable, alongside real growth in the trades that build and power data centres. Expect work and AI to be a central political issue.
- AI matches experts at most computer-based tasksAI capabilities · Medium confidence
- Organisations are redesigned around AIWork · Medium confidence
- A personal AI is the main way into digital servicesEveryday users · Medium confidence
- Clear displacement in clerical and support rolesJobs · Medium confidence
A different shape of work and daily life
Ten years is far enough that direction is clearer than timing. We expect AI to exceed human experts across most intellectual work, with people setting goals and carrying responsibility. If that happens, scientific progress speeds up sharply and the limits become experiments, trials and regulation.
Daily life changes with it: a capable personal assistant for everyone, more preventive healthcare, personal tutoring and autonomous transport. The mix of jobs looks very different, with many people in occupations that have no name today, and the skills that keep their value are the human ones: trust, care, taste, leadership and craft.
- AI exceeds human experts across most intellectual workAI capabilities · Medium confidence
- Most valuable computer-based work is done by AIWork · Medium confidence
- Everyone has a capable personal assistantEveryday users · Medium confidence
- The mix of jobs looks very differentJobs · Medium confidence
What this means for you
Learn to hand work to AI and check what comes back. The people who do well are those who know their field deeply and use AI well, not one or the other.
Pick the few uses that save real time or money and do them properly. Decide early how AI agents are given access, supervised and audited.
Skills that depend on physical presence, human contact, judgement and accountability hold their value longest. Expect to retrain more than once.