AI is changing work. What should improve for people?
An organisation introduces AI to allocate tasks and schedule work. Explore how worker participation, workload, autonomy and access to human review shape its implementation.
Affected people
Workers across relevant roles and shifts, including supervisors and people handling exceptions.
Intended improvement
Less repetitive work, better working arrangements, and stronger capabilities and autonomy; examine how gains and burdens are shared.
Participation & priorities
Define the tasks, roles and shifts in scope. Involve affected workers before rollout and agree which changes would be useful.
Commitment in practice
Make allocation decisions understandable and open to human review, with workers able to challenge harmful arrangements.
Proposed action Test adjustable schedules with workers; review workload and autonomy after rollout.
Potential evidence sources
Pre-deployment participation and review records, oversight arrangements, explanation and appeal records, workload data, and worker accounts.
Questions to examine
- whether affected people participated before deployment
- what human oversight exists in practice, not on paper
- how workload, pace and discretion changed for those working with it
- whether a decision can be explained to the person it affects
- whether appeal and correction exist, and whether they change outcomes
- how benefit and risk from the system are distributed
Interpretation and conflicting evidence
Compare organisational records with affected people’s accounts within the same scope and period. Record differences, evidence quality and the reasons for human judgement.
Open questions
Were affected workers and users involved before deployment, and can material automated decisions be explained, challenged and corrected?
Evidence gaps and provenance
For each item, retain the source, date, subject, scope, relevant passage and limitations. Confirm availability and coverage before drawing a conclusion.
- Examine indirect workload and behavioural effects.
- Examine distribution of productivity and financial benefits.
Responsible actors to consult about evidence: The deploying organisation for participation, oversight and appeal; named decision owners for material decisions.
Proposed action
Test adjustable schedules with workers; review workload and autonomy after rollout.
Responsible role
The transformation lead, with worker representatives and the manager responsible for human review.
Follow-up
Agree a review date before rollout. Compare workload, autonomy and access to review across relevant roles and shifts; record changes and unresolved concerns.
Improvement record
Record the agreed change, responsible role, review date and evidence to revisit. Keep unresolved concerns, the response to them and any revised action visible.
Check subsequent results
Agree a review date before rollout. Compare workload, autonomy and access to review across relevant roles and shifts; record changes and unresolved concerns.
Shared learning
Test one material automated decision end to end: participation before design, the explanation given, the override used, and the correction that followed.
Compare subsequent evidence with the declared baseline, scope and Framework version. Record what changed, what remains unresolved and what the method should learn.