UK Research and Innovation will begin combining eight grant-assessment reforms from October 2026, including staff-led triage, wider distributed peer review and a retrospective test of AI-assisted assessment. The initial AI exercise will compare model outputs with decisions already made; UKRI has not announced that an AI system will decide live awards.
What is changing in UKRI grant assessment?
UKRI says high application volumes are creating intense competition, low success rates and pressure on reviewers. Its 10 September announcement brings together eight measures intended to speed decisions while reducing unnecessary work.
The package includes more demand management for oversubscribed opportunities, larger and longer-term investments for some strategic programmes, greater freedom to tailor assessment methods and in-person verification for higher-risk awards.
Will AI decide who receives research funding?
Not in the trial as described. UKRI plans a retrospective exercise with the UK Metascience Unit: an AI-assisted process will assess past applications, and researchers will compare its conclusions with the decisions that human panels previously reached.
The announcement says human judgement will remain central. A retrospective test can expose agreement, bias, instability and failure modes without allowing the experimental output to determine a current applicant’s result.
What are desk rejection and distributed peer review?
Staff-led initial triage, sometimes called desk rejection, removes applications before full external review when they do not meet the relevant threshold or requirements. UKRI says it will test the approach rather than impose one identical method on every competition.
Distributed peer review asks members of the applicant community to assess other submissions. UKRI plans to expand it for suitable high-volume opportunities following research-council trials. The approach may spread workload and shorten queues, but it still needs safeguards for conflicts, consistency and confidentiality.
Can applicants use generative AI?
UKRI already has a separate generative-AI policy for applications and assessment. It sets responsibilities for applicants, reviewers and the funder, including transparency expectations and rules around sensitive application material.
The new assessment programme does not replace opportunity-specific instructions. Applicants should continue checking the published guidance for the scheme they are entering, because the method and evidence required may differ.
What should researchers watch next?
The important evidence will be whether the reforms produce faster decisions without weakening fairness, subject expertise or confidence. UKRI says it will monitor equality implications and use metascience research to evaluate effectiveness.
Its wider strategy aims to cut grant-processing times substantially by 2031. The October package is therefore best understood as a portfolio of experiments and operational changes, not a single switch to automated peer review.
Reporting basis: Ferdiox reviewed UKRI’s reform announcement and current generative-AI policy on 15 September 2026. This report distinguishes adopted process changes from experimental or retrospective work.
Illustrative image. Photo by Scott Graham on Unsplash