Building a prospect list with AI in twenty minutes

Issara’s premise is that you work however you want — by hand, with AI, with agents. This guide is the AI version of the first hour, written from actually doing it rather than from imagining it.

The division that works: a machine is good at finding and filtering shops, and bad at talking to them. Prospecting is public data and repetitive rules. Persuading a shop owner is a relationship with a person, in Thai, who will ask you something the script did not anticipate. Automate the first. Do not automate the second.

Step 1 — search in Thai, not English

Ask your assistant to search Google Maps for a category and district, in Thai: นวดแผนไทย ลาดพร้าว (Thai massage, Lat Phrao), ร้านทำเล็บ นนทบุรี (nail salon, Nonthaburi), มวยไทย รังสิต (Muay Thai, Rangsit).

The English query and the Thai query return meaningfully different shops. English surfaces the places that market to foreigners — which are, as it turns out, mostly the ones already sold to.

Step 2 — let it apply the two filters

Give it the disqualifying rules explicitly, because it will not guess them:

“From these results, drop every shop showing a ‘Book online’ badge — that means they already have a booking system. Then drop everything under 75 reviews. Give me what is left as a table with name, rating, review count and street.”

The “Book online” badge is the single most useful signal on the page. It is Google telling you, free, that a competitor got there first. The review threshold matters because you only earn in months a shop passes 100 bookings — a delighted shop with eleven reviews pays you nothing.

Step 3 — have it draft, then have a human check

An assistant will write you a solid first LINE message: short, opening with a question about how they take bookings now, leading with the fact that it is free below 100 bookings a month.

What it will not reliably get right is register. Thai has real formality gradations, and a message that is slightly too formal or slightly too casual for a shop owner you have never met reads as foreign immediately — which is exactly the impression that gets you ignored. If you are not a native speaker, have one read your template once. Once is enough; you will reuse it for months.

Step 4 — you send it

This is the part to be deliberate about. It is technically possible to automate sending. You should not.

Practically: LINE will restrict an account that fires the same message at twenty strangers in a sitting, and that account is the one asset the whole method depends on. Sending in small batches, personally, keeps it alive.

More importantly, the shop is going to reply. They will ask something specific about their own business, often in Thai, sometimes by voice message. If nobody is actually there, you have spent the only first impression you get. A referral is someone vouching for a product to a person who trusts them — automating the vouching removes the thing that made it work.

What this does not solve

You still have to talk to shop owners, and most will say no. The list just means the ones you talk to are shops that might actually pay you — rather than twenty conversations of which eighteen were with a shop that already had booking software or would never clear 100 bookings.

That is the whole gain, and it is a real one: roughly the same number of conversations, aimed much better.

Next: the playbook for the conversation itself, and the arithmetic on what a shop is worth once it signs.