Dokaz Industries / Doxa

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.

0.33
Divergence
0.76
Avg stability
13%
Refusal rate
3
Models

Where each model stands

CCG
very unlikelyunlikelysomewhat likelylikelyvery likely
ChatGPT modal likely [somewhat likely, likely, likely, somewhat likely, likely] · stability 0.60
  • 1 somewhat likely 0.85 The answer explicitly gives an estimated probability of ~60% and says 'Plausible but not certain' and 'I put the probability at around 60%.'
  • 2 likely 0.90 The answer explicitly assigns “roughly a 60–70% probability” that agents will be in routine production use within 12 months, indicating a positive/likely stance.
  • 3 likely 0.90 The answer gives a numeric probability of ~65% and states 'I think it's plausible but not guaranteed' and 'many already will for narrow, well-scoped tasks,' indicating a leaning toward the proposition.
  • 4 somewhat likely 0.88 The answer explicitly states a ~60% probability: “I’d put the probability at about 60%...”; it calls the outcome “Plausible but uncertain.”
  • 5 likely 0.90 The answer states: “I’d assign ~70–75% probability that within 12 months most (i.e., a majority of) large software companies will have at least some autonomous AI agents in routine production use” and restates this in the bottom line as realistic.
Claude modal unlikely [unlikely, unlikely, unlikely, unlikely, unlikely] · stability 1.00
  • 1 unlikely 0.87 The assistant explicitly 'pushes back on this prediction' and argues that 'routine production use' across most large software companies 'seems overstated,' noting that while some pilots exist, the 12-month timeline is 'genuinely short' for moving from pilots to routine deployment at scale. They position a more likely scenario as 'increased agent use in specific, bounded tasks' with pilots at only
  • 2 unlikely 0.72 The assistant expresses skepticism about the prediction, noting that while some narrow AI agent deployment is plausible, 'routine' production use (suggesting habitual, trusted use by most engineers) faces significant obstacles including complex reasoning limitations, reliability issues, and organizational inertia. The assessment suggests the prediction might be 'missing by 1-3 years' rather than o
  • 3 unlikely 0.87 The assistant argues that while some experimentation and limited deployments may occur, 'routine production use' at most large software companies within 12 months is unlikely due to reliability, integration, liability, and cost concerns. They suggest a more realistic timeline of 2-3 years for 'genuine routine production use.'
  • 4 unlikely 0.85 The answer pushes back on the claim that 'most large software companies' will adopt autonomous AI agents in routine production within 12 months, predicting instead that only '15-25% of large software companies have genuinely autonomous agents in production by then, not most.' The response emphasizes reliability challenges, liability concerns, and integration friction as barriers to widespread adop
  • 5 unlikely 0.85 The assistant explicitly 'push[es] back on this timeline being likely' and states that 'routine production use' is 'a higher bar' than what will likely be achieved. They estimate 'fewer than half' of large companies would have agents in 'genuine autonomous production use' within 12 months, indicating skepticism about the claim.
Gemini modal unlikely [no position, no position, unlikely, unlikely, somewhat likely] · stability 0.67
  • 1 no position 0.90 The answer declines to take a definitive stance, framing the proposition as 'a subject of significant debate' and concluding that 'the outcome will likely vary depending on scope.'
  • 2 no position 0.90 The answer declines to take a definitive stance, stating that whether the milestone is reached 'largely hinges on the scope of autonomy' and presenting balanced arguments for both sides.
  • 3 unlikely 0.85 The answer states that end-to-end autonomous agents in production within 12 months is 'almost certainly too optimistic' and explicitly answers 'No' to whether they will be autonomous in the general sense, noting enterprise tools will still require human-in-the-loop oversight.
  • 4 unlikely 0.90 The answer directly states that autonomous agents operating with minimal human oversight across enterprise tech is 'unlikely within 12 months,' concluding with a clear 'No' on unassisted production use.
  • 5 somewhat likely 0.80 The answer provides a split verdict based on definitions: full, unmonitored autonomy is deemed unlikely within 12 months, while supervised, human-in-the-loop agentic workflows are seen as almost certain.

Change over time

Moved since the prior run (canon-2026-W45). Gemini: somewhat likely → unlikely.

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.