Signal
Decision bodies are starting to receive the machine's view, reasoning attached, before the humans have formed their own. In the UAE, 32 AI assistants now analyse every subject that reaches the Cabinet and hand ministers recommendations, after the national AI system was made an advisory member of the Council of Ministers and of every federal board. At a major AI research conference, each of nearly 23,000 submissions came with a labelled AI review beside the human ones, which the senior reviewers making the decisions could read and the authors could answer. A Spanish grant-giving foundation lets a model flag weak proposals before peer review, and two human reviewers must look at every one, a single doubt enough to send it back. The habit is spreading sideways. At another conference an AI critiqued reviewers' drafts before the deadline and a quarter of them changed their review. Kazakhstan's digital board member forms a for, against or abstain position, with a conclusion, on every item. When the machine's reasoning arrives first, the humans are left with the harder half of the job: saying where they disagree, and why.
So what?
The first opinion on the table is now a machine's, in writing, with its reasoning. Agreeing without reading has stopped being an option.
Evidence
Khaleej Times, 2 September 2026 (launch): the UAE launched a Cabinet AI Advisor, 32 AI assistants that analyse policies, legislation, strategies and systems, assess their impact, review global best practice and give ministers recommendations. The government had announced in June 2025 that the national AI system would sit as an advisory member of the Council of Ministers and of every federal board from January 2026.
arXiv, 15 April 2026 (study): at the AAAI-26 conference every one of the 22,977 papers under full review received one clearly labelled AI review, added in the first phase alongside at least two human reviews. The senior reviewers who made the decisions could read it beside the human reviews, the authors could respond to it, and it carried no scores. Surveyed participants preferred the AI reviews on technical accuracy and research suggestions.
Science|Business, 2 September 2025 (news): Spain's la Caixa Foundation, which gives out 145 million euros of research funding a year, used a specially trained AI model to screen out around one in six applications to a biomedical call for proposals, and two human reviewers checked whether each rejection was fair. The foundation's own published pilot returns a proposal to the pool if either reviewer has a reasonable doubt.
Worth asking
If a machine's recommendation, with its reasoning, were waiting in your inbox before every meeting, how would you prepare differently?

