1. From prospect to accidental systems analyst
An independent developer publishes a web tool and receives a sales email from a stranger. The sender accurately describes a feature: turning photos of notes and work steps into the first draft of a manual. The offer is to build new products and an income stream. It is slightly flattering to discover that someone found the site.
The twist: the thing being sold is a system for finding sales prospects and writing personalized messages with AI. An AI-tool creator has become the target of AI-powered sales. Before deciding to buy, the developer starts examining how to build the interesting part. One email, one accidental engineering review.
2. How personal is "personalized"?
The opening lines mention the tool's actual function. Much of the remaining offer is broadly about earning income from one's own products. A single true detail can make an otherwise reusable message feel bespoke.
The site might have been found through search results, product directories, public pages or a published contact address. The actual discovery path is unknown. Nor can the email prove whether AI chose the recipient, wrote the wording or sent it. A message that looks AI-assisted is not evidence of full automation. Inventing unseen execution logs would turn an analysis into detective fiction.
3. What is the seller offering?
Published pages describe CredLayer as help with turning personal expertise into products and an online sales foundation, and SpeedSale as AI sales training for people with an existing product. The listed amounts are ¥30,000 per month for basic CredLayer and ¥50,000 including tax for SpeedSale. [1][2]
The seller describes a process of searching for prospects, studying each one, drafting a tailored message, having a person check it and press send, then responding or arranging a meeting. "50 messages in 30 minutes" is a sales claim, not an independently measured result. The actual search engine, AI model, contact source and mailing service are not identified.
The offer also describes resale rights and advertised commissions of ¥90,000–¥1,536,000 per sale. [2] It combines business setup, sales training and resale opportunities, not merely a secret app.
4. Seven ordinary jobs behind the magic
Define who might need the product; collect permitted public information; judge product fit; write an evidence-based proposal; check whether contact is lawful; obtain human approval; and record replies, opt-outs and outcomes.
Writing the email is only one job. Finding the right people is potentially the most valuable part. Sending 50 generic messages is fast distribution. Identifying five companies with verifiable reasons to be interested is research. A better accelerator does not fix a broken navigation system.
5. Finding possible buyers without reading minds
For a manual-making tool, useful public signals might be a factory discussing new-worker training, a logistics business describing standard work, or a retail chain seeking consistent onboarding. These are reasons to investigate, not proof of a desire to buy.
Store the organization, evidence URL, actual statement, date, inference, product fit and contact-permission status separately. "Hiring beginners" may be observable; "their manuals are terrible" is an unsupported guess. If an AI sends the latter as fact, it has become a fortune teller with a fake business degree.
Score candidates on evidence quality, useful product fit, realistic need and whether outreach is allowed. Check those scores against human judgments.
6. Build the discovery factory first
A prototype can collect organizations through permitted search sources, read public pages, separate facts from guesses with AI, and store results. Cloudflare Workers can schedule tasks, D1 can hold the records, and Gmail API can prepare unsent drafts for human review. [3][4][5]
These are possible tools for a self-built system, not the original sender's confirmed stack. Search fees, model costs, website terms, deduplication and error recovery still require work.
The first version does not even need a send button. Make a reliable map before building the cannon.
7. Sending 50 emails is not the same as selling
A promotional simulation assumes 50 contacts on each of 260 days: 13,000 contacts. At a hypothetical 2% sales rate and ¥90,000 commission, annual commissions would be ¥23.4 million. [2] The arithmetic works. The 2% conversion rate is an assumption, not verified performance.
Track relevant prospects, useful replies, meetings, actual purchases, complaints, unsubscribe requests, time and costs. A large reply rate means little if every reply says "never email me again."
Otherwise you get a wonderful dashboard and a domain reputation that is quietly on fire.
8. Legal and reputational guardrails are not optional
Japanese rules on advertising emails generally require prior consent or another legal basis, with defined exceptions such as some publicly disclosed business addresses. Finding an address online does not grant unlimited permission to send. Refusal notices and later stop requests matter. [6][7]
Gmail also has authentication and anti-spam requirements, including additional rules for high-volume senders. [8] Laws vary by recipient location. Avoid collecting irrelevant personal information, bypassing access restrictions or inventing a customer's problems.
An email can be deleted; the impression of a careless sender is harder to restore. Trust has no undo button.
9. Start with a zero-send experiment
First find 20–30 organizations from public information. Let a person label each relevant, irrelevant or lacking evidence. Then draft messages for promising leads and check the factual claims. Only after establishing a lawful basis should a tiny set of human-approved messages be tested.
Measure lead precision, evidence accuracy, draft edit rate, useful responses, opt-outs, complaints, costs and sales. If only three of 100 companies fit, fix discovery rather than sending speed.
The best part is that prospect discovery can be evaluated without contacting anyone.
10. The pitch delivered a blueprint
The email shows that publicly described tools may become material for another company's sales research. It does not prove paying customer demand or that AI actually sent the message.
The reusable idea is a system that finds people who might benefit from a product and shows why. A content factory makes things; a prospect-discovery factory finds where those things might matter.
One sales email arrived. The decision to buy remains open. Curiosity about building the system, however, closed the deal instantly.
References (8)
- SpeedSaleの公式説明（AIによる候補発見、相手の調査、個別文案、予約、税込5万円、2026-10-03更新） awabota.life
- CredLayerの公式説明（人が送信前確認、30分50件という主張、基本月額3万円、再販報酬、試算、2026-10-07更新） store.awabota.life
- Cloudflare Workersの定期実行 developers.cloudflare.com
- Cloudflare D1データベース developers.cloudflare.com
- Gmail APIの下書き作成 developers.google.com
- 日本の特定電子メール法（特に第3条） laws.e-gov.go.jp
- 消費者庁による事前同意規制の例外と拒否表示の説明 caa.go.jp
- Gmail送信者向け要件・FAQ support.google.com
