6 min read
This is an illustrative story. Maya is a composite based on common patterns among solo makers who sell online, not a real individual, and her numbers are realistic estimates rather than reported results.
It is 9:40 on a Tuesday night. Maya has clay under her fingernails, a kiln cooling in the garage, and 23 unread messages in her shop inbox. Eleven of them are some version of the same question: “Can you make this mug in green, with my dog’s name on it, by December 12?” She loves these orders. They are her best-paying work. But answering them one by one, each with a slightly different price, glaze note, and shipping caveat, has quietly become a second job.
Six weeks later, Maya answers the same flood in about half the time, and her custom orders are up. Here is exactly how she did it, and why the setup is so refreshingly small.
The problem: custom orders were eating her studio hours
Maya sells handmade ceramics through her online shop and a small newsletter. About a third of her revenue comes from custom work: personalized mugs, matching dinnerware sets, pet memorial bowls. Every custom request triggered the same routine:
- Work out a price based on size, glaze, and personalization.
- Check whether the timeline fit her firing schedule.
- Explain that handmade glazes vary, so the color will not match a screen exactly.
- Ask for the details she always forgot to ask for the first time (handle side, font, spelling of the name).
She estimated roughly 7 hours a week going into custom-order email in the fall, climbing toward 10 as holiday season approached. Worse, slow replies cost her sales. Buyers who wait two days often simply order elsewhere.
The fix: one reference doc and Gemini in Gmail
Maya already used Gmail for her shop email, and her Google Workspace plan included Gemini, Google’s AI assistant. According to Google’s help page on collaborating with Gemini in Gmail, it can draft and refine replies right inside the email you are writing. Google has also been rolling out new Gemini tools aimed at small business owners. She did not need a new app. She needed to teach the one she had about her business.
Step one: a “custom order rulebook” in Google Docs
Maya spent one Sunday afternoon writing a two-page document with everything she normally typed from memory:
- Her price table (base mug price, add-ons for names, sets of four, rush fees).
- Her current lead time and the last date she could accept holiday orders.
- Her standard glaze disclaimer, written in her own friendly voice.
- A checklist of details every custom order needs.
- Three past replies she was proud of, as tone examples.
This was the breakthrough. The AI was never the hard part. Writing down what lived in her head was.
Step two: draft every reply against the rulebook
When a custom request arrives, Maya opens Gemini alongside the email, points it to her rulebook, and asks for a reply that quotes the right price, confirms the timeline, includes the glaze note, and asks for any missing details. She reads the draft, adjusts a sentence or two, and sends.
She keeps one firm rule: she checks every number by hand before sending. AI can misread a table or invent a figure that sounds right, so prices and dates always get a human look.
Step three: a saved prompt for the tricky ones
For requests that fall outside the rulebook, like a 40-piece wedding order, Maya uses a second prompt: “Summarize what this customer is asking for, list what I would need to know to quote it, and draft a warm reply asking those questions.” She still prices those jobs herself, but she no longer stares at a blank screen.
The results after six weeks
Here is what changed for Maya, using illustrative but realistic figures for a maker at her scale:
- Custom-order email time: down from about 7 hours a week to roughly 3.5.
- Typical reply time: from “next day or two” to the same evening.
- Back-and-forth per order: from three or four emails to usually two, because the first reply now asks for every detail at once.
- Custom orders accepted: up noticeably in the weeks before the holiday cutoff, which Maya credits mostly to faster replies.
The extra hours went straight back to the wheel. That is the real win: AI did not make her pottery, it gave her time to make more of it.
What made this work (and what nearly did not)
Maya’s first attempt went badly. She asked Gemini to “reply to this customer” without any rulebook, and it produced a cheerful, generic email with a made-up price. That is the most common mistake we see, and it is why we keep repeating that context beats clever prompts.
Three things made the second attempt succeed:
- A single source of truth. One document, updated whenever prices or lead times change. When she raised her rush fee, she edited one line instead of retraining herself.
- Real examples of her voice. Pasting in three past replies did more for tone than any instruction like “be friendly.”
- A human check on numbers. Fast drafts, careful sends.
How to copy Maya’s setup this week
You do not need to be a potter. This pattern works for any business that answers the same type of question with small variations: photographers quoting packages, bakers pricing custom cakes, tutors explaining schedules.
- Spend one hour writing your rulebook: prices, timelines, policies, and the questions you always need answered.
- Add two or three of your best past replies as tone examples.
- Use whichever AI assistant lives inside your email (Gemini in Gmail, Copilot in Outlook) or a saved project in ChatGPT or Claude.
- Draft every reply against the rulebook, then check numbers and dates yourself.
- Update the rulebook the moment anything changes.
Our story of a solo home baker who cut cake inquiries from 6 hours a week to under 2 shows the same idea from a different angle, adding an order form to collect details up front. And if you want product photos and listing graphics to match your faster replies, our Canva Craft review covers Canva’s new image model.
Holiday season is the perfect time to try this, since custom requests peak just when your hours are scarcest. What is the one question your customers ask you over and over? Drop it in the comments, and it might become the next story we tell.
