Agent Experience Standards for Public Service Systems

A specification for how autonomous agents interact with government and public service infrastructure.

Version 1.0 Draft. This standard defines the interface layer between autonomous AI agents and public service systems. It establishes requirements for machine-readable policy, structured action surfaces, human handoff protocols, audit accountability, context delivery, and the public value principles that govern them all.

The standard applies to any system where an autonomous agent takes actions on behalf of or in service to the public: benefits administration, permitting, licensing, case management, service delivery, and related domains.

Guide Structure

Part 1: Policy Logic

Policy as versioned, machine-readable interfaces. Covers canonical representation, version control, and reasoning attachments.

Part 2: Action Surfaces

Defined surfaces for agent interaction with public service systems. Covers action types, context requirements, outcome rules, and escalation conditions.

Part 3: Handoff Protocols

Protocols for transitions between agents and caseworkers. Covers triggers, handoff packages, caseworker response paths, and context restoration.

Part 4: Audit Trails

Accountability through complete reasoning records. Covers reasoning history, version anchoring, temporal records, and rights review.

Part 5: Context Delivery

How systems deliver context to agents. Covers structured data access, machine documentation, stable patterns, and consent flows.

Part 6: Public Value

Design constraints that govern interpretation and application of the technical standards. Covers equity of access, clarity, protection of judgment, and continuous review.

Standards Domains

Cross-cutting requirements across accessibility, security, transparency, interoperability, user experience, and compliance.

About This Standard

The Agent Experience Standards address a gap in current AI governance: the lack of specific, implementable technical standards for how autonomous agents interact with public infrastructure.

  • Parts 1–3 define the interface layer: how policy is represented, how agents act, and how handoffs work.
  • Parts 4–5 define the accountability layer: how actions are recorded and how context is delivered.
  • Part 6 defines the value layer: the principles that govern interpretation and application of the technical requirements.
  • The Standards Domains define cross-cutting requirements that apply across all six parts.

This is a living document. It will be revised as implementations surface new requirements and as the technology and policy landscape evolves.