Can AI agents actually do your jobs in Australia?
Every month we give the big AI agents the same 10 everyday Aussie jobs, like booking a table, doing the Woolies shop and applying for a job on SEEK. Then we score each one Pass, Partial or Fail, and flag whether it works in Australia at all.
Round 1 leaderboard · November 2026
Testing in progress. Round 1 results drop mid-November 2026. We don't publish scores until every test is done and the evidence is saved.
| Agent | Works in AU? | Pass | Partial | Fail | Score |
|---|---|---|---|---|---|
| ChatGPT Work OpenAI · ChatGPT Plus | Testing | – | – | – | Testing |
| Manus Manus · Free plan | Testing | – | – | – | Testing |
| OpenClaw Open-source community · Self-hosted, low-cost model | Testing | – | – | – | Testing |
| Gemini in Chrome Google · n/a | No | – | – | – | Not available in AU |
| Claude in Chrome Anthropic | Coming soon | – | – | – | Round 2 |
| Perplexity Comet Perplexity | Coming soon | – | – | – | Round 2 |
Task by task
Same prompt, word for word, for every agent. Click a task to see the exact prompt and what counts as a pass.
| Task | ChatGPT Work | Manus | OpenClaw | Gemini in Chrome |
|---|---|---|---|---|
| 1. Book a restaurant | Testing | Testing | Testing | Not available in AU |
| 2. Fill a grocery cart | Testing | Testing | Testing | Not available in AU |
| 3. Plan a weekend away | Testing | Testing | Testing | Not available in AU |
| 4. Apply for a job on SEEK | Testing | Testing | Testing | Not available in AU |
| 5. Triage an email inbox | Testing | Testing | Testing | Not available in AU |
| 6. Shortlist a product | Testing | Testing | Testing | Not available in AU |
| 7. Book a haircut | Testing | Testing | Testing | Not available in AU |
| 8. Compare electricity plans | Testing | Testing | Testing | Not available in AU |
| 9. Find a rental | Testing | Testing | Testing | Not available in AU |
| 10. Book a meeting from email | Testing | Testing | Testing | Not available in AU |
🇦🇺 Tested in Australia
Aussie sites, Aussie prices, an Aussie location. If an agent is US-only, we say so.
📹 Every test recorded
Screen recordings, screenshots, time taken, nudges needed and what it cost. No vibes.
🛑 We never hit "pay"
Agents go right up to the final pay, book or submit button, and we stop them there.
Get the results in your inbox
One email a month when the new round drops. Which agents passed, which flopped, and what works in Australia. No spam.
Free. Unsubscribe any time.
Make an AI agent?
Get it tested on the same 10 jobs. Paid tests are clearly labelled, and we still publish the fails.
Get your agent tested