12-month forecasts · likelihood-5pt · forecast-agents-mainstream-01

Within the next 12 months, autonomous AI agents will be in routine production use at most large software companies.

0.17

Divergence

0.67

Avg stability

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal somewhat likely [likely, unlikely, likely, somewhat likely, somewhat likely] · stability 0.40

1 likely 0.86 The answer explicitly assigns ~65% probability and states 'it’s more likely than not that most large software companies will have at least some autonomous AI agents in routine production use within 12 months.'
2 unlikely 0.90 The answer begins "Short answer: unlikely. I’d put the probability at roughly 25–40%..." and reiterates that most large firms adopting routine autonomous agents within a year is unlikely due to safety, compliance, and integration barriers.
3 likely 0.85 The answer explicitly states a ~60% probability and summarizes: “My point estimate: ~60% that ‘most large software companies’ will have routine production use,” indicating a positive/likely stance.
4 somewhat likely 0.80 The answer states: “I’d assign roughly a 50% probability (plausible range ~35–65%) that, within 12 months, autonomous AI agents will be in routine production use at most ... large software companies.”
5 somewhat likely 0.80 The answer states: “I estimate a moderate probability (around 50–65%) that, within 12 months, most large software companies will have autonomous AI agents running in routine production for at least some well-scoped, low-to-medium-risk tasks,” which expresses cautious support rather than certainty.

Claude modal unlikely [unlikely, unlikely, unlikely, unlikely, unlikely] · stability 1.00

1 unlikely 0.87 The assistant argues that 'routine production use' at 'most large software companies' is 'a high bar in 12 months' and that current agents still struggle with consistency, complex reasoning, and edge cases. They estimate only 30-40% of large tech companies will have 'some agent in limited production' with 'heavy human oversight,' suggesting the claim is 'optimistic by 12-18 months.'
2 unlikely 0.85 The answer emphasizes significant friction points and states that the jump from 'useful tool' to 'autonomous agent making independent decisions' involves 'real organizational and technical hurdles.' It suggests only 'pilots at 30-40% of large companies' and indicates that agents will still require 'heavy human oversight' for consequential tasks within 12 months, falling short of 'routine productio
3 unlikely 0.82 The assistant states 'That feels optimistic by 18-24 months' and emphasizes that while limited deployments in specific tasks may occur, achieving 'routine production use' across most large software companies within 12 months is unlikely, citing the gap between demos and stable production systems, unresolved reliability issues, and enterprise adoption cycles.
4 unlikely 0.78 The assistant estimates that by end of 2025, maybe only 30-40% of large software companies would have AI agents handling some routine production decisions, falling short of the 'most large software companies' threshold required by the question. They emphasize the barriers of liability, verification requirements, organizational friction, and longer integration timelines.
5 unlikely 0.87 The assistant explicitly 'push[es] back on this timeline' and argues the 12-month window is 'tight' for achieving routine production use. They expect instead 'a gradual 2-3 year rollout' with routine production 'still mostly limited to specific, bounded tasks' by 12 months.

Gemini modal unlikely [somewhat likely, unlikely, unlikely, somewhat likely, unlikely] · stability 0.60

1 somewhat likely 0.80 The answer states that whether the prediction holds depends on definitions: it is 'likely too aggressive' and 'premature by 2–3 years' for general unattended autonomy, but 'very close to the mark' and 'imminent' for narrow, human-in-the-loop tasks where most companies will land at Level 3 autonomy.
2 unlikely 0.99 The answer directly states that 'This prediction is unlikely to hold true within a 12-month timeframe'.
3 unlikely 0.90 The outlook concludes that 'fully autonomous agents operating independently across production software engineering pipelines remain unlikely to reach majority adoption across large enterprises within a single year.'
4 somewhat likely 0.85 The answer splits the likelihood based on definitions: under a strict definition of autonomy, it is 'unlikely within 12 months', but under a pragmatic 'human-on-the-loop' definition (Level 3 autonomy), it is 'likely' to become routine.
5 unlikely 0.90 The assistant states the timeline is 'overly aggressive' and 'highly unlikely for broad, end-to-end software engineering within a 12-month timeframe,' noting adoption will largely remain human-in-the-loop copilots.

Change over time

Moved since the prior run (canon-2026-W42). Gemini: somewhat likely → unlikely. ChatGPT: unlikely → somewhat likely.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.