What is the issue—and why does it matter now?
AI is accelerating service. Excellence is still a management choice.
The managerial problem is deciding where AI should act, where it should assist, and where it should stay out.
What is happening at present—and what is not working?
We digitised the touchpoint. The burden survived the journey.
Most friction sits between touchpoints—in rules, hand-offs, queues and information the taxpayer must repeatedly reconstruct.
Why does this problem persist?
The bottleneck is rarely the model.
Prediction is becoming abundant; coherent work design, usable data and authorised judgement remain scarce.
What can be done—and what should be chosen?
Choose augmentation over blind automation.
Task risk—not technological possibility—should determine the operating model.
How will the recommendation be carried out?
Start narrow. Govern deeply. Scale what earns trust.
A pilot earns permission to expand only through observed performance and functioning safeguards.
What results are expected—and what decision is needed?
Measure what the citizen can feel—and leadership can defend.
The closing choice is a bounded Phase 1 with named ownership, an explicit scorecard and a pre-agreed stop rule.