Standards Domains

The Agent Experience Standards operate across six domains. Each domain represents a category of requirements that cuts across all six parts of the standard. Compliance requires meeting requirements in all domains simultaneously.

Accessibility

Agent interfaces must be accessible to the full range of people who interact with public services — including people with disabilities, limited English proficiency, low digital literacy, and those using assistive technologies.

• All agent-facing interfaces that produce public-facing outputs must comply with WCAG 2.1 AA standards at minimum.
• Plain-language requirements apply to all human-readable outputs, including reasoning explanations, status updates, and handoff summaries.
• Multi-language support must be provided where the served population includes significant non-English-speaking communities.
• Alternative channels (phone, in-person, paper) must remain available and provide equivalent service levels.

Security and Privacy

Agent interactions with public service systems involve sensitive personal data and consequential determinations. Security and privacy are structural requirements.

• All data in transit between agents, systems, and users must be encrypted using current standards.
• Agent identity and authorization must be verified at every interaction, at session initiation and throughout.
• Personal data access must follow the principle of minimum necessary — agents receive only the data required for the specific action being performed.
• Data retention must comply with applicable privacy regulations. Personal data must be purgeable when retention periods expire, except for audit records required for accountability.
• Security incidents involving agent systems must be reported, investigated, and disclosed according to defined incident response protocols.

Transparency

Agent systems operating in public service contexts must be transparent to oversight bodies, the public, and affected individuals.

• The use of AI agents in any public service process must be publicly disclosed, including the scope of agent involvement and the types of determinations agents participate in.
• Aggregate performance data — processing times, error rates, escalation rates, outcome distributions — must be published on a regular schedule.
• The policy rules that agents apply must be publicly available in both machine-readable and human-readable forms.
• Audit mechanisms must be available to authorized oversight bodies, including legislative auditors, inspectors general, and designated external reviewers.

Interoperability

Public services span multiple agencies, jurisdictions, and systems. Agent interfaces must support interoperability so agents can operate across system boundaries without losing context or accountability.

• Action schemas, outcome types, and data formats must follow published standards to enable cross-system compatibility.
• Handoff packages must be portable — a handoff from one agency's system must be readable and actionable by another agency's caseworkers or agents.
• Policy version identifiers must be globally unique and resolvable across systems.
• Audit trails must be linkable across system boundaries, so a case spanning multiple agencies maintains a complete, unified record.

User Experience

The person at the center of a public service interaction must experience a coherent, respectful process — regardless of whether agents, humans, or a combination are handling their case.

• People must never be required to re-submit information they have already provided, regardless of whether the recipient is an agent or a caseworker.
• Status updates must be proactive, timely, and delivered through the person's preferred channel.
• Error messages must be actionable — telling the person what went wrong, what they can do, and who to contact for help.
• The overall experience must be designed as a unified journey across all system interactions.

Compliance and Governance

Agent systems operating in public service contexts must comply with all applicable laws, regulations, and institutional policies — and must be governed by structures that ensure ongoing compliance.

• Each deploying agency must designate an accountable officer responsible for the agent system's compliance with these standards.
• Regular compliance audits must be conducted by parties independent of the team that built or operates the agent system.
• Non-compliance findings must trigger defined remediation processes with deadlines and escalation paths.
• Changes to agent systems — including model updates, policy rule changes, and schema modifications — must go through a defined change management process that includes compliance review.
• Vendors and contractors involved in building or operating agent systems must be contractually bound to these standards and subject to compliance verification.