Agentic Surfaces
Trust Calibration How to read a pattern →
4.5 Frontier

Communicating the level of automation

Naming how much of an action a machine decides, in a plain label a user can grasp at a glance. A user reading the label knows whether a person or a machine settled the outcome they're about to rely on.

01

The impact of agents

As agents capable of varying degrees of automation are deployed across government services, the degree will differ from one service and action to the next, and the tools to build them are now widely available. Without a shared frame for naming that degree, the same agent behavior reads as helpful efficiency to one user and as threatening opacity to another. Neither can tell how much a human decides.

02

What must be verified

Government needs a user to be able to tell how much of a given action a human decides versus a machine, in terms they can grasp without technical literacy. The label then addresses both misuse (over-reliance) and disuse (under-use) rather than leaving reliance to guesswork. The service or agency deploying the agent must assign that label and keep it accurate as the automation changes.

03

Protecting access

An abstract scheme like 'Level 2 supervised automation' means nothing to a stressed or low-literacy user. A label can be technically accurate and still leave someone's reliance miscalibrated if they can't parse it. That means they may follow a machine decision believing a person checked it first, or avoid a service altogether that a machine could have handled safely.

Keeping the path open

  • Use concrete, action-oriented language: 'a person will check this before it's final'.
  • Test comprehension with low digital literacy, cognitive disability, and limited-English cohorts.
  • Give an example of what each level means for this specific service, rather than in the abstract.
04

Response surface

Automation Label

How much of this action was automated is stated as one of three plain labels, shown before the user consents to it.

Preview the same interaction at each automation label
Recalculate your rebate for the new quarterAI-assisted
What this label meansThe AI helps you prepare; every substantive step is yours
Where a person sitsA person makes the decision, with your prepared material in front of them
What the agent will doRead the new quarter’s rate table, recalculate your rebate, and lodge the updated amount
Data it will accessYour current claim · household record · this quarter’s rate table. No other data is accessed.

The label is set by what the system does, checked against the oversight records. A service cannot claim ‘AI-recommended’ while its reviewers approve at a rate no different from ‘AI-decided.’
05

Maturity

  1. Established

    For automation-level labeling in adjacent domains (automotive autonomy levels, open-banking consent), where the response is proven.

  2. Frontier Headline

    As a public-facing scheme for government AI services, which remains proposed rather than operated.

06

Precedents

Parasuraman and Riley on use, misuse, disuse, and abuse. The paper names four ways people and automation combine: use is voluntary activation or disengagement, misuse is over-reliance producing monitoring failures and decision bias, disuse is neglect commonly caused by alarms that fire falsely, and abuse is automating a function without regard for the consequences for the person left holding what remains. Abuse is a designer's and manager's failure rather than an operator's, and the paper finds it promotes the other two. Misuse and disuse are both addressable through clear automation-level signaling.

SAE J3016 as a communication device. Beyond its technical function, the levels framework succeeded as a communication tool: 'Level 2' or 'Level 4' conveys meaningful information to non-experts. 'Level 1: AI assists, you decide' is more comprehensible than 'supervised machine-learning-augmented decision support'. A numbered scale survives translation into ordinary speech.

APP 1.7-1.9, automated-decision transparency. The Privacy and Other Legislation Amendment Act 2024 legislates an automation-level disclosure: a privacy policy must state which decisions are made solely by a computer program and which are substantially assisted by one. The trigger, decisions that 'significantly affect the rights or interests of an individual', is broader than the GDPR Article 22 test of solely automated processing. The duty is up-front policy disclosure, and not a per-decision statement of reasons.

07

What carries over to agent use

High transferability. Government services suit a simple, public-facing automation-level vocabulary. A three-level scheme is likely sufficient:

  • "AI-assisted": a human makes the decision; the AI helps gather and organize information.
  • "AI-recommended": the AI proposes a decision; a human reviews and approves it.
  • "AI-decided": the AI decides within defined rules; a human is available for review on request.

The open-banking consent pattern transfers well: before any agent action with consequences, show the user what the agent will do, what data it will access, and offer explicit consent or decline.

08

Where things go wrong

The failure mode is an automated decision presented as if a human exercised judgment, hiding the absence of meaningful review. Labeling determinations honestly as 'AI-decided' makes that absence legible to the people affected. The label itself can be gamed downward: a service can mark an action 'AI-recommended' when a reviewer only glances at the output, borrowing the credibility of the lighter label without doing the review it implies.

09

Sources

6 references Global · UK · Australia
Primary frameworks

The instrument, the operating deployment, or the official record itself.

Reporting & analysis

Writing about the subject rather than the framework itself, including vendor commentary.