Most small delivery shops in Ashland do not have a marketing department, and the person writing the Instagram captions is often the same person packing orders and answering the phone. Lately, a lot of owners have been turning to AI tools to speed up the writing, and the early results are mixed. Generic prompts produce generic copy, and generic copy is risky in a category where advertising rules are tight. One approach that has helped is starting with midjourney prompts and similar ready-made prompt libraries, then adapting them to a local, compliant voice instead of asking a chatbot to “write something cool about weed.”
Why a generic prompt fails for cannabis delivery
A prompt like “write a fun post about our new gummies” tends to return copy that makes health claims, uses slang your customers may not trust, or promises effects no one can verify. Cannabis marketing is restricted in most places, and platforms add their own rules on top of state law. The problem is not the AI tool. The problem is that the prompt gave it no boundaries, no audience, and no sense of your shop.
A usable prompt for a delivery business needs four things: the product category, the customer you are speaking to, the channel, and the lines the copy must not cross. When those are spelled out, the output becomes a draft you can edit in ten minutes instead of a rewrite you dread.
A basic structure you can reuse
Here is the structure that works best for our shop. Treat it as a template and fill in the brackets with your own details.
- Role: “You are writing for a licensed cannabis delivery service in Ashland, Oregon. Your tone is friendly, plain, and local.”
- Audience: “Adults 21 and over who already buy cannabis and want convenient delivery. Do not address minors.”
- Task: “Write a 60-word product description for [product], covering only the listed facts: [weight, strain type, flavor, packaging].”
- Constraints: “Do not make health, medical, or therapeutic claims. Do not promise effects. Do not mention prices or discounts. Do not use emojis that could appeal to children. Include the standard age notice.”
- Format: “Return one paragraph and two alternative headlines.”
Notice that the prompt supplies the facts. Asking the model to invent details is where errors creep in, so give it the product sheet and ask it only to rephrase.
Use cases that save real time
Product descriptions for the online menu
Your menu probably has dozens of SKUs, and many of them differ only by flavor or package size. Create one prompt that takes the product sheet as input and returns a consistent description. Consistency matters because customers compare items side by side, and mismatched tone looks careless.
Delivery window updates
Customers want to know when their order is coming, and they want it without a wall of text. A short prompt can turn your dispatch notes into a friendly message: “Your order is out for delivery and should arrive between 4 and 5 p.m. Please have your ID ready.” The ID reminder is not optional for compliance, so build it into the template rather than trusting yourself to remember it every time.
Local event and community posts
Ashland has a strong arts and outdoor scene, and local events give you topics that generic national posts cannot match. A prompt that asks for a short post about a local trail weekend or a farmers market, with a clear instruction to avoid consumption suggestions, can produce a warm, community-focused draft. Be careful with event promotion, since some venues and sponsors have their own cannabis restrictions. Check before you post.
Email reminders for returning customers
Reorder reminders work best when they are plain and respectful. Ask for three subject line options under 40 characters, each one neutral in tone. Avoid anything that sounds like pressure or urgency tied to a health need.
Building your own prompt library
The most valuable thing you can do is keep a running file of prompts that worked. Each entry should include the prompt text, the channel it is meant for, the date you last tested it, and a note about what you changed by hand. Over a few months this becomes an internal asset that a new staff member can use on day one.
If you do not want to build every prompt from scratch, look at what marketplaces for AI prompts offer. Some sellers publish prompts tested for specific image and text tools, and you can adapt those for your store. The key is to read each one critically. A prompt written for a fashion brand will not know about age gating, and it may push language that is wrong for your product. Our shop has found it useful to treat any purchased or borrowed prompt as a first draft that needs a compliance pass before it goes live. For a sense of how these libraries are organized, a prompt library built for product and menu copy is a reasonable place to see what structured entries look like.
A compliance checklist before anything goes live
Before publishing any AI-assisted content, run through this list. It takes less time than a single customer complaint.
- Does the copy make any health, medical, or effect claims? If yes, cut them.
- Is the age notice present where your state or platform requires it?
- Would the post be appropriate if a minor saw it? Remove characters, cartoons, or language that appeals to young people.
- Are prices, discounts, or promotions described in a way your local rules allow?
- Did a human read it aloud once? Awkward phrasing is easier to catch out loud.
- Does it accurately describe the product? Compare every claim to the product sheet.
Keep in mind that rules differ by state and by platform, and they change. Check current requirements with your licensing authority or a qualified attorney rather than relying on any AI tool to know them.
Where AI helps and where it does not
AI is good at rewriting, reformatting, and producing options. It is poor at knowing your customers personally. It does not know that your regulars in the Siskiyou foothills want straightforward information about delivery windows, or that your team prefers a dry, understated voice. Those details come from you. The best workflow is a human who knows the shop, a prompt that carries the rules, and a final read before publishing.
There is also a trust question. Customers who buy from a local delivery service are often choosing a business they know by name. Copy that sounds like it came from a template can quietly erode that. Use AI to get to a first draft faster, then add one specific detail only your shop would know, such as the name of the driver who has been doing deliveries since you opened or the route you take up the hill on a snowy morning.
Getting started this week
You do not need a large project to begin. Pick one recurring task, such as the delivery window message, and write a prompt with the four elements above. Test it on five real orders. Edit what needs editing, save the version that worked, and move to the next task. Within a month you will have a small set of reliable prompts that cover most of your routine writing.
Start small, keep your rules visible, and never let a tool decide what your business is allowed to say. That discipline is what turns AI from a risk into a reliable part of running a local delivery shop.
Leave a Reply