OECD’s discussion of skill needs in the age of AI emphasises changes in task and skill composition and the need for training to respond.[2] ZHICE therefore proposes organising learning around actual decision gaps rather than tool catalogues. A person may operate the system yet not understand an anomaly’s significance. Conversely, an experienced field practitioner may struggle to communicate the basis of a judgement to a central analytics team.

Effective transfer combines what happened, why the judgement was made and when the same action should not be repeated. Steps without reasoning encourage mechanical imitation; expert conclusions without an evidence path are difficult to review. Extract verified learning cases from completed work, retaining counterexamples and boundaries. Relevant specialists should confirm applicability so an accidental success does not become a general rule.

Use a progression of observation, supervised practice, independent work and exception review. Each stage should correspond to observable capability rather than accumulated hours. Allocate time for mentors explicitly instead of treating teaching as an invisible addition to specialist workloads. Where real tasks are rare, clearly labelled simulations can support judgement practice, but applicability still needs testing in actual work. Simulated performance does not establish full field capability.

A knowledge library needs operational ownership. Policy updates, equipment replacement or process changes can alter the validity of old cases. Maintainers should receive change notifications and decide whether to update, restrict or withdraw material. Managers should examine which tasks use the knowledge, whether repeated requests decline and whether cases are applied incorrectly. The volume of content is not the outcome. People must explain the evidence, recognise boundaries and escalate when appropriate.

The ILO’s just transition guidelines include social dialogue in transition policy processes, involving governments, workers and employers’ organisations.[3] This is an international policy reference, not a substitute for local employment requirements. For enterprise management, it highlights that work change involves how people understand responsibility, learning opportunities and future roles, not only a revised process diagram. Engagement belongs during design, rather than after a solution is complete.

Engage people around specific tasks instead of a general debate about whether technology is good. Ask different roles which information is hard to obtain, which handoffs create rework and which automated recommendations are hardest to assess, then test whether the new design addresses those issues. State uncertainty where resources or role arrangements remain undecided rather than make unsupported assurances. Credible communication rests on clear facts, boundaries and follow-up, not an optimistic answer to every question.

Pilot feedback must be able to change the design. Maintain a structured route for technical defects, responsibility conflicts and inadequate capability support, with owners and feedback dates. Performance arrangements should reward accurate reporting of risks and boundaries rather than treat every pause or escalation as inefficiency. If people fear that recording exceptions will harm their assessment, operational data may become distorted and the organisation loses an early-warning channel.

Participation should lead to visible outcomes: a simpler handoff, clarified authority, funded practice or a deferred proposal because evidence is weak. Explain why suggestions are not adopted and what would prompt reconsideration. This turns participation from a one-off meeting into continuing improvement, while helping management distinguish legitimate work constraints from information gaps instead of attributing every disagreement to resistance to change.