AXHUB LECTURE · INTRO L3

A shop owner's first AI
— start with review replies

Time about 30 minutesWhat you need one chatbot + 3 recent reviews of your shopFor self-employed and small-business owners

The reason to pick review replies for your first AI is simple. It repeats every day, it has rules, and it shows when it falls behind. As one owner running an online shop alone put it — to start, handing off one task you repeated today is enough.

1Why start with review replies (3 min)

Touch big things like customer service or ordering first and you'll fail.

Review replies are different — the input is clear (the review text), there are rules (thanks, empathy, guidance), and mistakes can be undone (check before you post).

It has every condition of a task that's easy to automate.

Check: how many reviews are backed up right now without a reply?

2Make your shop's 3-line rule (7 min)

Recall how you usually reply and write it in three lines.

These three lines are your shop's "reply manual" and the instruction you'll give the AI.

Practice 1 — examples by trade (pick and adjust)
[Cafe] These are our cafe's review reply rules. (1) Thank them (2) Pick up specifically on the menu item mentioned (3) One line about the seasonal new item. For a negative review, apologize first, no excuses. Polite, two or three sentences.

[Restaurant] (1) Thanks (2) Pick up on the mentioned dish or wait situation (3) One line about reservations. For hygiene complaints, apology + only the action taken.

[Hair salon] (1) Thanks (2) Pick up on the service name (3) Note the next visit timing. Mention the stylist's name only if it's in the review.

You can just swap the words to fit your shop. The key is to always include the "negative review rule."

Check: do the rules have all three — (1) order (2) what to avoid (3) tone?

3Run it on 3 real reviews (10 min)

Paste in 3 recent reviews of your shop, as they are.

Two praise reviews and one disappointing review is best.

Practice 2
[the 3-line rule from step 2]

Below are three real reviews of my shop. Write a reply draft for each.
[review 1]
[review 2]
[review 3 — disappointing review]

If a draft is off, fix it with "this sentence isn't our shop's voice, make it like this: ___." Two or three rounds and it improves.

Check: does the disappointing-review reply start with an apology and carry no excuses?

4Before posting, a person checks last (5 min)

Don't post the draft as-is.

One rule — for high-star reviews you may post the draft as-is, but for low-star reviews the owner always refines and posts it personally.

Across the successful cases, this way of having a person handle only the exceptions split the results by more than double.

Check: did you set "the owner's-eyes threshold" (at or below how many stars)?

5Next expansion + free support (5 min)

Once review replies are second nature, the next candidates are set — a regulars re-visit notice, a new-item intro, FAQ answers.

It's all the same method: 3-line rule → real material → owner check.

If you want to learn more, free government support is wide open. The AI online courses at the Small Enterprise Knowledge Learning Center, the AI win-win collaboration training with curricula by trade, and the Innovation Small Business AI Adoption Support program that connects training to on-site coaching and commercialization funding (around 2,000 people in 2026). The recruiting rounds keep changing, so it's best to check the current one at Sobizz24 (sbiz24.kr).

Check: did you decide on the next repeated task to hand off?

When it doesn't work

The tone doesn't sound like our shopPaste in two past replies you wrote yourself and like, and say "this voice is the standard." Faster than ten lines of rules.
It invents a menu item or situation not in the reviewAdd one line to the rules — "Don't mention anything not in the review." If it still happens, point to that sentence and rerun.
I'm worried about customer informationBefore pasting, remove the customer's name, nickname, and order number. Also add to the rules that no customer info goes into the reply.

Next lecture → L4. Meeting notes, reports, email — the office worker's three repeated tasks · Full contents

Sources: government small-business AI support tracks (Knowledge Learning Center, win-win collaboration training, Innovation Small Business AI Adoption Support — per official notices, verified 2026-07) · Stanford Enterprise AI Playbook (review-exceptions-only +71% vs +30%) · solo-operator reviews. Full sources are in the axhub.net case library. The practice prompts are example lines to follow along.