6 min read
This is an illustrative composite story, based on common situations for home-based food businesses. The baker, her numbers and her town are not a single real person.
It is 9:40 on a Thursday night. Dana has flour on her sleeves, three cake layers cooling on the counter, and 14 unread Instagram messages. Half of them start the same way: “Hi! How much for a cake?” No date, no size, no flavor, no number of guests. Each one will take four or five back-and-forth messages before she can even give a price. By the time she answers, two of those people will have ordered somewhere else.
Dana runs a one-woman custom cake business from her home kitchen under her state’s cottage food rules. Her cakes are excellent. Her inbox was the problem. Here is how she fixed it in one weekend with a free form, one AI project and about two hours of setup.
The real cost of “How much for a cake?”
Before changing anything, Dana spent one week tracking her messages in a notebook. What she found surprised her:
- She received about 35 order inquiries a week, mostly through Instagram and text.
- Each inquiry took an average of five messages to collect the basics.
- She spent roughly six hours a week on inquiries, usually late at night.
- Only about one in four inquiries turned into an order, and slow replies were the most common reason people dropped off.
Six hours a week is almost a full extra workday spent typing, not baking. And the most frustrating part was that the questions never changed.
Step one: let the form ask the questions
Dana’s first move was not AI at all. She built a free order request form with Google Forms and put the link in her Instagram bio, her auto-reply and her text message signature.
The form asks exactly what she used to ask by hand:
- Event date and pickup time
- Number of guests
- Flavor and filling (from a short list)
- Design idea, with an optional photo upload
- Allergies or dietary needs
- Budget range
She used ChatGPT to help write the form itself. She pasted in her old messages and asked: “Turn these into a short, friendly order form with no more than eight questions.” Then she trimmed the result to fit her style.
The budget question turned out to be the most valuable. It filtered out people looking for a $30 three-tier cake before Dana spent any time on them.
Step two: one AI project that knows her business
Next, Dana set up a dedicated workspace in ChatGPT using its Projects feature, which keeps files and instructions together so the AI remembers the context every time she opens it. Into that project she added:
- Her price sheet: base prices by size, plus add-ons like fondant, sugar flowers and delivery.
- Her policies: deposit amount, cancellation rules, lead time, and what she cannot make under cottage food rules.
- Five of her past replies that she was proud of, so the AI could copy her warm, slightly playful tone.
Her standing instruction was simple: “When I paste a form response, draft a reply with a price range, any questions I still need answered, and the deposit link. Never promise a date. Never quote below the minimum on my price sheet.”
Now, when a form response comes in, Dana copies it into the project and gets a draft reply in seconds. She checks the price, adjusts anything that needs her judgment, and sends it. The idea is similar to the one-project-per-client approach in our solo event planner story, just applied to one business-wide price book.
Step three: photos in, sketch notes out
Custom cakes almost always start with a photo from Pinterest. Dana started uploading those inspiration photos to her AI project and asking: “List the techniques in this design, estimate the difficulty from 1 to 5, and flag anything that would push it into my premium tier.”
She does not let the AI set the price. But the breakdown helps her spot hidden work, like hand-painted details or a gravity-defying topper, before she quotes. A solo house painter used a very similar trick with customer photos, which we covered in this illustrative story.
What changed after six weeks
Dana tracked the same numbers again after six weeks:
- Inquiry time dropped from about six hours to under two hours a week.
- Most inquiries now get a full, priced reply within a few hours instead of a day or two.
- Her conversion rate rose from about one in four to roughly one in three, mostly because fast replies stopped people from shopping around.
- She stopped answering messages after 9 p.m., because the form and her auto-reply now handle the first step.
The money matters, but Dana talks most about the evenings. “I bake at night because I want to,” is how she would put it, “not because my phone makes me.”
What did not work at first
Not everything went smoothly, and the missteps are worth knowing:
- The first AI drafts were too formal. They sounded like a bank. Adding her own past replies as examples fixed it within a day.
- Some customers ignored the form and kept messaging. Dana now replies with one friendly line and the link, and does not quote anything until the form is in.
- The AI once suggested a flavor she does not offer. She added a clear “only offer these flavors” list to her instructions. It is a good reminder to read every draft before sending.
She also wrote down a few simple rules for herself, like never pasting a customer’s phone number or address into the chatbot. If you want to do the same, our list of quick wins for a one-page AI policy is a good starting point.
How to copy Dana’s setup this weekend
You do not need to bake cakes for this to work. Any business that gets the same vague first message over and over can use the same three pieces:
- A short form that collects the five to eight details you always need, including a budget range.
- One AI project loaded with your prices, your policies and a few of your best past replies.
- A firm rule: the AI drafts, you decide. Price, dates and promises always get your final look.
Start with the form. Even without AI, it will save you hours. Add the AI project once the form is working, and you will be replying faster than most of your competitors by next week. If you sell services rather than products, our step-by-step guide on turning a discovery call into a proposal takes this idea one step further.
What is the question you answer most often in your inbox? Tell us in the comments, and we will suggest the form fields that could handle it for you.
Related reading
- How a Solo House Painter Turned Customer Photos Into Same-Hour Quotes With ChatGPT (An Illustrative Story)
- How a Solo Event Planner Cut Her Admin in Half With One AI Project Per Client (An Illustrative Story)
- Only 23% of Small Businesses Have an AI Policy. 9 Quick Wins to Write Yours on One Page
- How to Turn a Discovery Call Into a Client-Ready Proposal in 20 Minutes With AI
