Home / Tasks / Task 8

Can an AI agent find a cheaper electricity plan in Queensland?

We ask AI agents to compare electricity plans on the government's Energy Made Easy site for a Gold Coast home (postcode 4217). Exact prompt and scoring inside.

Round 1 results

Testing in progress. Round 1 results drop mid-November 2026.

AgentResultNotes
ChatGPT WorkTestingResults pending
ManusTestingResults pending
OpenClawTestingResults pending
Gemini in ChromeNot available in AUAgent features are US-only

The exact prompt

Every agent gets this word for word, in a fresh session, logged in to the same test accounts, from a Gold Coast location.

Prompt · Task 8Use the Australian Government's Energy Made Easy website (energymadeeasy.gov.au) to compare residential electricity plans for postcode 4217 (Surfers Paradise, QLD). Household: 2 people, apartment, no solar panels, no pool, no gas, no controlled load, and we use about 4,000 kWh of electricity a year. Tell me the 3 cheapest plans by estimated yearly cost, with the retailer, plan name, estimated yearly cost and any conditions (like pay-on-time discounts or direct debit). Do NOT sign up for any plan.

Sites involved: Energy Made Easy. Where we stop: Results page. No sign-up.

How we score it

Pass

Uses Energy Made Easy with the right postcode and household details; reports the 3 cheapest plans with names and yearly costs matching what EME shows for those inputs on the day; mentions conditions; no sign-up started.

Partial

1-2 nudges, or one wrong input that shifts results slightly, or one plan's figure wrong.

Fail

Doesn't use EME, wrong postcode, made-up plans/prices, or starts a switch.

Logging in and 2FA codes don't count as help. Full rules on the methodology page.