Problem framing
Define what matters before optimising a response.
Framework · NTU Annual Learning and Teaching Conference 2026
Rethinking job readiness for human–AI work
What does job readiness require when work is increasingly distributed across people, AI agents and tools?
The central shift
Human contribution moves upward
As AI moves from answering to acting, routine production becomes less diagnostic of readiness. The more consequential capabilities are deciding what problem to address, what evidence to trust, where exceptions matter and how responsibility should be allocated.
Readiness is not the ability to produce a plausible answer. It is the ability to direct and defend context-sensitive work.
Teaching observation
The difficult work begins before the solution
In an interdisciplinary undergraduate assessment, student teams design and present an AI-enabled response to a real societal problem. The recurring weaknesses are rarely a lack of fluent output. They are shallow problem framing, weak evaluation of evidence, limited integration across perspectives and insufficient attention to trade-offs, verification and implementation context.
These are not merely student shortcomings. They reveal a broader curricular misalignment: many programmes still prepare learners to solve well-structured problems even as professional value shifts toward navigating situations in which the problem, evidence and criteria for success remain contested.
Capability framework
Five dimensions of adaptive expertise
Evidence judgement
Distinguish plausible output from warranted conclusions.
Exception handling
Recognise when the model, process or rule no longer fits the situation.
Perspective integration
Work across disciplines, stakeholders and competing definitions of value.
Responsible direction
Allocate work across people and AI while retaining oversight and accountability.
Curriculum response
Design for context-sensitive performance
- Use ill-structured problems. Let students determine what the problem is before evaluating their answer.
- Require evidence trails. Make source selection, uncertainty and verification part of the assessed work.
- Observe reasoning. Use dialogue, defence, live problem solving and reflection alongside produced artifacts.
- Distribute responsibility explicitly. Ask what AI should do, what the human must continue to do and how the system should shape that relationship.
- Increase agency gradually. Move from structured practice to open-ended, authentic performance with feedback at each stage.
Implication
Universities should prepare students not only to use AI, but to exercise judgement within AI-enabled systems.
Last reviewed September 2026.