AppStore2031

Possible future 1 of 6 · W01

AI helps; humans stay in charge

AI becomes a normal workmate, but people still make important decisions and remain responsible.

Forecast target
31 Jul 2031
Evidence cut-off
2 Aug 2026
Edition
2031-2026-08-02
Status
Working forecast

This is a scenario, not a prediction or probability. It is a deliberately testable possible condition used to see how product demand changes.

How this future develops

Now

Where we are

AI can already draft, analyse and organise work, but it still makes mistakes and often needs checking.

World shift

What would change

AI improves and spreads, but not enough for organisations to trust it alone with important decisions.

2031

What life could feel like

Most people still have jobs. AI prepares and checks the work; a named person decides, explains and fixes what goes wrong.

Why the apps would change

The most useful apps help people work faster while making responsibility, approval and recovery clear.

The condition the research tested

If cross-domain reliability remains brittle or costly to assure through 2030, and institutions keep a named human responsible for consequential action, machines mainly prepare, check and coordinate work. Paid employment remains the main income, status and time allocator in July 2031, while physical, climate, care and education constraints keep gains local and uneven.

Causal chain

  1. Reliability and recovery remain costly.
  2. Audit, liability and security keep authority scoped.
  3. Organisations redesign tasks around supervised assistance.
  4. Wages and payroll-linked protection remain central.
  5. Care, teaching, maintenance, climate adaptation and security retain human duty.
  6. Work structures time and status, while contribution and belonging also depend on family, community and civic institutions.
  7. Relational AI sometimes bridges to people and sometimes substitutes for reciprocal contact.
  8. Physical and fiscal limits keep outcomes regionally uneven.

What we had to assume

  • Independent audit and liability remain influential.
  • Employers and educators preserve supervised expertise pipelines in some capable systems.
  • No single compound shock destroys managed interdependence everywhere.

What nobody knows yet

  • Whether constrained delegation reflects capability or institutional choice.
  • Whether complementarity improves bargaining or intensifies surveillance.
  • Whether gains finance care, education and climate adaptation.
  • Whether relational systems bridge to people or displace reciprocity for particular groups.

Evidence used to frame this world

observation · W01-E01

At the cutoff, agent progress coexists with short dependable horizons, jagged transfer and self-validation gaps.

Source records: AC-06, AC-09, src-international-ai-safety-report-2026, src-metr-long-tasks-2025

Important limit: Evaluations overrepresent digital benchmark environments and do not prove failure everywhere.

observation · W01-E02

Exposure measures do not establish realised displacement, and bounded studies show mixed productivity.

Source records: EWF-04, EWF-07, SC-WORK-01

Important limit: Occupation exposure and bounded studies do not settle economy-wide employment.

projection · W01-E03

Growing compute demand meets grid, transformer, water and local-permission constraints.

Source records: PC-S01, PC-S02, E05

Important limit: Projected totals are not commissioned local capacity.

policy-intent · W01-E04

AI rules, public compute, workforce programmes and assurance plans point toward supervised adoption.

Source records: AC-25, AC-30, AC-33, AC-36

Important limit: Plans and laws do not prove delivery, enforcement or equity.

inference · W01-E05

Costly review, rescue, security and liability preserve accountable human roles and supervised entry paths.

Source records: EWF-07, src-ilo-exposure-indicators-2026, src-who-multimodal-health-guidance-2024

Important limit: This causal link is inferred, not an observed 2031 result.

scenario-condition · W01-E06

Independent tests do not jointly cross long-task, low-rescue, matched-cost and broad-diffusion gates by 2030.

Source records: src-oecd-ai-trajectories-2026, src-metr-long-tasks-2025

Important limit: The conjunction is a scenario classifier, not a forecast.

unknown · W01-E07

Net wages, work intensity, entry careers, informal work, unpaid care and relational-system effects remain unsettled.

Source records: EWF-04, EWF-05, SC-AI02

Important limit: No global causal series joins these outcomes.

Open the complete source register →

Forecast apps most affected by this future

RemedySwitch

Challenge a machine-made decision and reach a qualified human who can stop it, explain it and put the result right.

CaseCairn

Stop being bounced between services: one qualified human carries your case until the next service genuinely acts.

DutyPool

Call an independent qualified duty officer to pause a high-stakes machine action and organise a fair repair.

HumanHarbour

Step back from an embedded synthetic relationship without losing your records, support or route to a real person.

ShiftWitness

See whether a machine can finish the whole shift, including every rescue, delay and repair.

ClaimTide

Make old proof expire the moment a machine, workplace or rescue plan changes in a meaningful way.

HumanLift Ledger

Count the human effort hidden behind an automated claim and give workers power to challenge it.

CareKeel

No silent cancellation: keep the agreed care minimum moving through an outage.

LoadCovenant

Approve a huge new load without quietly making homes and public services pay the price.

RecoveryLoom

Practise the failure nobody expects—before your team has to recover a real hospital, grid or delivery system.

Mendstep

Fix real homes and local services, get paid while learning, and climb into skilled work one safe repair at a time.