Curriculum-aware task routing is a proposal that AI systems track a junior colleague’s skill progression and give them tasks at the edge of what they can do, including tasks the system could have done itself. It comes from a Google DeepMind paper on delegation, where it appears as a design goal.
The answer, in one line
A proposal from Google DeepMind authors that AI systems track a junior's skill progression and allocate tasks at the boundary of their expanding skill set, including work the system could do itself.
Definition#
Curriculum-aware task routing: systems that “track the skill progression of junior team members” and “strategically allocate tasks that sit at the boundary of their expanding skill set”. The phrase and the quotations are Tomašev, Franklin and Osindero’s (“Intelligent AI Delegation”, arXiv:2602.11865, 12 February 2026, Section 5.6).
The problem it answers#
The passage sits in a section the authors title “Risk of De-skilling”. A system that delegates to people on grounds of speed and cost will give routine work to the AI and leave the hard residue to humans, which removes the practice that builds skill. Curriculum-aware routing reverses the allocation rule for juniors: the system sometimes hands over work it could do, because doing it is how the junior grows. Graded entry.
What has and has not been shown#
The paper is a conceptual framework posted as a preprint. In the text read for this page it reports no experiments, datasets or results, and the routing idea is stated as something the authors say should be developed, without a built or tested system. No limitations are listed for it. The evidence for its premise, that practice builds skill and that removing practice erodes it, comes from elsewhere in this research, for example the colonoscopy study, and none of it tests this mechanism.
The cost it asks an organisation to accept#
This section is interpretation. Giving a junior a task the AI could finish faster means a slower result today for a more capable colleague later. Whether an organisation will do that depends on who bears the delay and who sees the gain, and a system that tracks progression also needs a measure of progression that is not output. That second requirement is the hard one, and it connects to assessing capability rather than output.
Key sources
- Tomašev, N., Franklin, M. and Osindero, S. (2026). Intelligent AI Delegation. arXiv:2602.11865, 12 February 2026. Preprint. Graded entry.
- Budzyn, K. et al. (2025). Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy. The Lancet Gastroenterology and Hepatology, 10(10), 896-903. Graded entry.
Related SuperSkills research#
See never-skilling, how juniors become senior and the delegation boundary map. The phrase is DeepMind authors’, not a SuperSkills coinage.
About this research#
Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.
Explainer · SS-2026-420 · Graded against the published rubric · 1 working paper
Hirji, R. (2026). What is curriculum-aware task routing?. The SuperSkills evidence base, SS-2026-420. https://thesuperskills.com/research/what-is-curriculum-aware-task-routing. Last reviewed 12 October 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work