Agent-readiness pilot
Can AI agents find your product, sign up and finish real work? We test it, and we check the result ourselves.
Email Miles to start
The problem
AI agents now pick, sign up for and use developer tools on their own. Two things go wrong, and you rarely see either:
- Agents say “done” when the job isn’t. The run log says “account created” or “deployed”. Your database, inbox or API says otherwise.
- Sign-up and verification flows stop agents. Bot challenges, email codes, phone checks, SSO-only logins and keys that only a dashboard can issue stall an agent. The person who sent it never learns why.
Every stalled agent is a sign-up you never knew you lost.
What we test
- One product, up to 3 real workflows you choose, each run end to end: discover, sign up, then a core task (for example, create a project and make the first API call).
- 2 agent setups (for example, a browser agent and a coding agent using your API, CLI or MCP server), 3 repeat runs each: 18 runs in total, so we can tell flaky from broken.
- Discoverability: llms.txt, docs as markdown, API and OpenAPI, MCP, robots rules.
- Sign-up friction: bot challenges, email or phone verification, SSO-only logins, and whether an agent can get a key or token without a human.
How we verify
We don’t take the agent’s word for it. Each run’s end state is checked separately through your API, a database query or a test inbox. A negative control is built in, so a false “success” shows up as a failure. Every run is recorded with traces: steps, requests, screenshots and timings.
What you get
- A reproducible report: pass or fail on the verified end state for each workflow and agent, with traces and the scripts so your team can rerun it.
- A failure list: where each run stalled or wrongly reported success, and why.
- Prioritised fixes, ranked by impact on agent completion and effort.
- One re-test after you ship a fix, with before-and-after results.
Timeline: 10 business days from access and deposit
Days 1–2: kickoff, pick workflows, set up test accounts. Days 3–7: runs and verification. Days 8–9: report. Day 10: report delivered, with a walkthrough by call or written Q&A. Re-test any time within 30 days of the report.
Price and terms
- A$4,500 fixed (about US$3,130 at 6 October 2026 rates). No hourly billing.
- 50% deposit (A$2,250) to start; balance on delivery of the report.
- Full fee credited against a Humanless implementation (A$25,000–50,000) if signed within 60 days of the report.
- We don’t solve captchas, bypass security controls, load-test, or touch real customer data. Test accounts are cleaned up afterwards.
What we need from you
- A named contact and a short kickoff (call or email).
- Up to 3 workflows, and what “done” looks like for each.
- One way to check the end state: a read-only API key, a database query or a log or webhook view.
- Permission to create test accounts, and whether to test your bot protection as it is (we report where it stops agents) or through a test bypass.
Next step
Reply with your product and up to three workflows you want tested. We’ll send a short order form and the deposit invoice the same day. A free sample audit is available on request.