Writing Cannabis Delivery Copy with AI Prompts: A Practical Guide for Michigan Teams

Written by

in

Running a cannabis delivery operation in Michigan means writing a steady stream of copy: menu descriptions, order status texts, FAQ pages, driver instructions, and social posts that have to stay within state advertising rules. Most small teams hand this work to whoever has a free afternoon, and the result is uneven. Some operators have started looking at prompts for content creators as a shortcut, and the idea is worth a closer look, as long as you understand what separates a useful prompt from a generic one.

Why generic AI output causes problems in this niche

Ask a general-purpose AI tool to write a product description for a gummy or a pre-roll and you will usually get enthusiastic language about relaxation, wellness, and how the product will change your evening. That is exactly the kind of copy a cannabis business should not publish. Health and effect claims are tightly restricted, and marketing that appeals to people under 21 can create real trouble for a licensed business.

The fix is not to avoid AI entirely. It is to give the tool tighter instructions than most people bother to write. A prompt that specifies the audience (adults 21 and older in Michigan), the format (a 40-word menu blurb), the banned content (no medical claims, no effect promises, no references to youth-oriented themes), and the factual inputs (strain type, THC and CBD percentages from your lab results, package size) will produce something you can actually use after a light edit.

What a working prompt contains

Across most delivery-related tasks, a dependable prompt has five parts:

  • Role: who the writer is, such as a copywriter for a licensed retailer.
  • Audience: who will read it, including the age requirement.
  • Source facts: the exact data from your product record, pasted in, so the model does not invent potency numbers or terpene claims.
  • Constraints: word limits, tone, prohibited phrases, and required disclaimers.
  • Output format: a headline, a body, or a plain text message under a set character count.

If a prompt is missing one of these, the output usually tells you which one. When a draft wanders into vague praise or unsupported claims, the constraint section almost always needs another line.

Compliance guardrails to build into every prompt

Before you use any template in production, have your compliance lead or attorney review the language. Michigan’s rules for licensed marihuana businesses cover advertising content, audience targeting, and packaging, and the details matter. Build these guardrails into your prompt library:

  • Never state or imply that a product treats, cures, or prevents any condition.
  • Pull potency and weight directly from the certificate of analysis, never from memory or a previous listing.
  • Include the age requirement in any customer-facing content that could reach a general audience.
  • Avoid cartoon imagery references, candy-style names in promotional copy, and anything that appeals to minors.
  • Keep a log of approved outputs so you can show how content was reviewed.

AI does not carry responsibility for what you publish. Your business does. Treat every generated draft as a first pass from a junior writer who has never read your state’s rules.

Practical use cases for delivery operators

Menu descriptions

Menus change weekly as harvests and inventory shift. A prompt that takes a product record and returns a short, factual description saves significant time. Feed it the category, the lab-verified numbers, the flavor notes from your buyers, and the banned-phrase list. Then spot-check every output against the source record before it goes live.

Order status and delivery texts

Customers want to know when their order is leaving the shop, when the driver is nearby, and what to have ready at the door, including ID. These messages are short and repetitive, which makes them good candidates for templated prompts. The prompt should specify the exact information to include and forbid any promotional language in transactional messages, which keeps them clean and easier to defend. To go deeper, explore The marketplace for AI prompts that actually work.

FAQ pages

Questions about delivery windows, minimum order sizes, service areas, and payment methods are common. Ask the model to draft answers from a list of verified policies you provide. Do not let it guess at policies. If a policy is not in your input, the prompt should instruct it to write a placeholder rather than invent an answer.

Driver and staff training notes

Internal documents can use AI too. A prompt can turn a messy list of procedures into a one-page checklist for new drivers covering ID verification, handling refused orders, and documentation. Because these documents are internal, the tone can be plainer, but accuracy still matters, so have a manager review them against current procedures.

Building a review workflow

The biggest mistake is treating prompts as finished products. A prompt is a tool that needs maintenance. A reasonable workflow looks like this:

  1. Write or adapt a prompt for one specific task.
  2. Run it on three to five real product records, including edge cases such as high-THC items and products with incomplete data.
  3. Have a second person check each output against the source record and the compliance list.
  4. Save the approved version with a date and the name of the reviewer.
  5. Revisit the prompt whenever regulations, product lines, or brand voice change.

This process adds a little time up front and removes much of the risk later. It also gives you a library that new staff can use without rebuilding everything from scratch.

Keeping your brand voice distinct

Michigan’s cannabis delivery market is crowded, and many shops sound identical. Use your prompts to protect a voice that reflects your actual business. If your team is known for plain, practical language, say so in the role section and provide two or three sample sentences you have already approved. If your strength is local knowledge, such as knowing which neighborhoods you serve best or how your drivers handle rural routes, put that context into the source facts. Generic output sounds generic. Specific input produces copy that sounds like you.

What AI cannot replace

No prompt will know that a particular batch smells sharper than last month, or that your customers keep asking about a product you stopped carrying. Those observations come from staff who handle the product and talk to buyers. Use AI to reduce the time spent on repetitive writing, and keep human judgment in charge of accuracy, tone, and compliance. The goal is to free your team to focus on service, which is what keeps delivery customers coming back.

Getting started this week

Pick one task that eats the most time, such as menu descriptions or order texts. Write a prompt with all five components described above. Test it on real records, have it reviewed, and save the approved version. Once that works, expand to the next task. A small, well-tested library will do more for your operation than a long list of untested templates.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *