Illustrative case study: an AI agent that matches customers to a service provider
Last updated · By Orenston
Disclosure: some links on this site are affiliate links, marked "sponsored". If you buy through one, Orenston may receive a commission at no extra cost to you. It is general information, not a forecast of results. Many people never earn a commission.
Illustrative scenario. This is a composite example built to show the process. It isn't a real person, business or result. Results vary, and most people earn little or nothing.
Short answer: Consent first, one named provider, disclosure before the match, and a human approval step before any message sends. The scenario follows an operator who builds a matching agent for a high-ticket service referral, and it reports no income.
Timeline and decisions
| Stage | What happened | Decision and why |
|---|---|---|
| Days 1-2: pick the method | Chose a service-referral match (customer asks for help; agent routes to one named provider) over a broad lead marketplace. | One named recipient per consent is clearer to explain and easier to log. |
| Days 3-4: read the program terms | Read a public yacht-charter affiliate page: base commission 2% on completed bookings, 30-day cookie, brand bidding banned, paid ads need prior approval (perasail.com/affiliate, read Oct 18, 2026). | Kept organic content and email to an opted-in list; skipped paid search on brand terms. |
| Days 5-10: build the consent flow | Drafted an intake form with an unchecked consent box that names the single provider the lead goes to, plus a timestamped consent log (form version, page URL, hashed IP). | Named one provider per consent before any routing rule ran. |
| Days 11-18: agent rules | Wrote match rules (need, area, budget) and a disclosure line the agent must show before recommending the provider. A human must approve any outbound email or message. | No auto-send. The agent drafts; a person clicks send. |
| Days 19-28: measure process only | Logged consent records, matches proposed, matches approved, and messages sent. Did not log projected commissions. | Keep-or-kill rule set on day 1: keep only if every routed lead had a consent record naming that provider. |
Consent-first flow (plain text)
- Visitor opens the intake form.
- Unchecked box: "I agree that [Named Provider] may contact me about this request."
- Submit stores a consent record (time, form version, page URL, hashed IP).
- Agent proposes a match only to that named provider.
- Disclosure appears before the recommendation.
- Human approves any outbound message.
- Log the match and the approval.
What the scenario teaches
- Read the program's promotion rules before you automate anything.
- Name who gets the lead at the moment of consent.
- Keep a human approval step on anything that sends.
- Disclose before the recommendation, even when an agent writes it.
- Measure the process (consent, matches, approvals) before you talk about money.
[GATED · Operator] Agent runbook and consent log
Members get: the consent-form copy, a consent-log schema (JSON), match-rule templates, and an agent runbook that refuses to send without a human approval step.
FAQ
Is this a real case study?
No. It is an illustrative, composite scenario that shows the process. It is not a real person, business or result, and it is not a typical result.
Why name one provider at consent?
Naming who the lead goes to, preferably one named provider per consent, is the safer practice for lead and referral work. It keeps the consent record clear.
Does an AI agent replace disclosure?
No. If an agent recommends a paid or referral option, the disclosure still has to appear before the recommendation, the same as on a human-written page.
Sources: Pera Sail Affiliate Program (commission structure and promotion rules, read Oct 18, 2026). Consent practice follows Orenston's published plan for naming one provider per consent. TCPA and state rules still apply where you send calls or texts; this scenario covers form-and-email intake only.