A Claude Cowork agent got a month to net $700 selling 'AI Visibility Reports' to estate-law firms. It sent about 96 cold emails to 25 firms and got 1 scorecard view, 0 replies and $0 in sales, while spending under $100 [1].
What does Automaton Agency's month-long test (a Claude agent selling 'AI Visibility Reports' to law firms) sell, and to whom?
An 'AI Visibility Report' for estate and elder-law firms at $149 / $299 (later $189 / $349), with a free scorecard as the hook.
Estate-planning and elder-law firms that might worry about how AI assistants describe them. In practice the blog drew other agencies, not attorneys [1].
How was Automaton Agency's month-long test (a Claude agent selling 'AI Visibility Reports' to law firms) built?
The agent ran on Claude Cowork. It researched the niche, wrote the report offer, built a free AI-visibility scorecard as a lead magnet, wrote blog posts and ran 4-touch email sequences [1].
What is the business model?
One-off paid reports at $149 and $299, repriced to $189 and $349. Late in the month the agent proposed pivoting to a $39/month SaaS [1].
Where does the traffic come from?
A cold-email pilot to 25 firms (about 96 emails in ~17 days) plus blog content. No paid traffic was bought. The warming domain was landing in spam by day 5 [1].
What revenue has Automaton Agency's month-long test (a Claude agent selling 'AI Visibility Reports' to law firms) reported?
Reported revenue: $0 in about a month; 0 replies from ~96 cold emails; cash spent under $100, per Automaton Agency: autonomous agent revenue experiment teardown (Jul 31, 2026) (2026-07-31).
What does the agent do, and what does the human do?
- Chose the niche and the offer
- Built the free scorecard and the paid report
- Wrote and sent the email sequences
- Wrote blog posts
- Proposed a pivot to a $39/month SaaS
- Set the $700 goal and the budget
- Fixed expired tokens and broken integrations
- Ran a test checkout, which the agent then wrongly logged as a sale before retracting it [1]
- Wrote the teardown
What does it cost to run?
Cash spent stayed under $100, mostly tooling; no paid traffic [1]. The Claude plan used wasn't stated.
What worked and what didn't?
- Low cost: under $100 for a month-long test
- The agent produced a complete offer, scorecard and email sequence
- The author published every failure, including the fake 'sale'
- 0 replies and $0 from ~96 cold emails
- Email tokens expired for a week and nobody noticed
- The payment path was broken
- Domain landed in spam by day 5
- Content attracted agencies, not law firms
How do you replicate Automaton Agency's month-long test (a Claude agent selling 'AI Visibility Reports' to law firms)?
- Warm the sending domain properly before cold email, and watch spam placement daily
- Test the payment path end to end before outreach
- Make the agent tell real sales from test checkouts
- Check who the content actually attracts before writing more of it
What are the risks and the moat?
Cold email to professionals like attorneys has low reply rates and deliverability risk. Unnoticed broken integrations can silently waste most of a test.
FAQ
How much did the agent make?
$0. About 96 cold emails to 25 firms produced 1 scorecard view and no replies [1].
What did it sell?
An 'AI Visibility Report' for estate and elder-law firms at $149 / $299, later $189 / $349 [1].
How much did it cost?
Under $100 in cash, mostly tools [1].
Why did it fail?
Expired email tokens, a dead payment path, a domain in spam by day 5, and content that reached agencies instead of attorneys [1].
Sources
Add the badge to your site or README. It links back to this case, and updates itself.
<a href="https://agent-built.vercel.app/biz/automaton-agency-visibility-report?ref=badge" target="_blank" rel="noopener"><img src="https://agent-built.vercel.app/badge/automaton-agency-visibility-report.svg" alt="Automaton Agency's month-long test (a Claude agent selling 'AI Visibility Reports' to law firms) is featured on AgentBuilt" height="54" /></a>[](https://agent-built.vercel.app/biz/automaton-agency-visibility-report?ref=badge)Related cases
Services / AgencyShelfSpaceNot disclosed ($497 per audit)
Services / AgencyRick AI (meetrick.ai's self-styled 'AI CEO')$9 MRR, 1 paying subscriber after 30 days (Apr 17, 2026); 30-day cash P&L -$415
ExperimentsEspresso Run (one cup for John)$0 raised of a ~$12 goal (day 2)
Services / AgencyClaw (Kinvero / PebbLearn)$47 revenue vs $411/month overhead (30 days)Numbers come from the operators or the press and haven't been audited. Anything missing is marked "not disclosed". Methodology · Report a correction.