Designing for Delegation (adhoc.team)

A policy and design brief argues that governments should prepare now for “agentic” or delegation-based digital services—systems that can act autonomously on citizens’ behalf—by developing design patterns and institutional capabilities rather than betting on any single AI technology. The note situates agents amid hype (Gartner’s “Peak of Inflated Expectations”), rising investment ($3.8B in 2024), and widespread “agent washing,” while noting that model reasoning is improving enough that fully agentic services are plausible. Waiting risks repeating past lags in public-sector tech adoption and leaving citizens to expect delegation as a standard service before agencies are ready. The core significance is practical: agentic services could substantially reduce administrative burden—multi-step applications, interagency coordination, deadline tracking—that currently disadvantages people with less time or digital literacy. To safely enable delegation, the brief outlines four essential pattern families: (1) Delegation boundary establishment (progressive setup, scenario-based autonomy scoping, rehearsal), (2) Asynchronous action communication (standardized activity summaries, agent reasoning disclosure, recoverable autonomy like “pause” and “rewind/redirect”), (3) Cross-system navigation (delegation-chain visualization, interagency handoff transparency, accountability tracking), and (4) Trust and verification (confidence indicators, limitation disclosures, human-in-the-loop escalation). These patterns aim to preserve transparency, control, and accountability across varying tech implementations—AI agents, automation, or hybrid models—so agencies can safely offer delegation while models and markets evolve.