Running a cannabis delivery business in Spokane means juggling a lot of writing: menu descriptions, order confirmations, driver updates, review replies, and the occasional apology when a route runs late. Many owners have started experimenting with AI writing tools to save time, and many have found that the first answers they get are generic, wordy, or occasionally wrong. The fix is rarely a better tool. It is usually a better prompt. Some operators now look to marketplaces offering premium prompt templates as a starting point, then adapt them to local rules and their own brand voice.
Why Generic Prompts Fail for Delivery Businesses
A prompt like “write a product description for a cannabis gummy” will produce something fluent and almost useless. It does not know your customer base, your state’s advertising restrictions, your packaging, or the tone your regulars expect. In a regulated category, a vague prompt can also produce copy that makes health claims, implies medical benefits, or appeals to minors. Those are not small problems. They can create licensing headaches and damage trust with customers who expect straightforward information.
A prompt that works for a delivery operation usually contains four things:
- A clear role for the AI, such as “you are writing for a licensed Washington retailer that offers delivery within city limits.”
- Constraints that are stated, not assumed, such as no medical claims, no references to effects, and no content aimed at people under 21.
- The exact output format you need, such as a 40-word description, three bullet points, or a two-sentence text message.
- A sample of the voice you want, pasted directly into the prompt.
Building a Prompt Library for Your Shop
The most useful thing a Spokane delivery team can do is keep a shared document of prompts that have already produced good results. Treat it like a standard operating procedure. Each entry should include the task, the prompt text, a sample of a good output, and a note about what to check before publishing. Over a few months, that document becomes more valuable than any single tool subscription.
Start with the tasks that repeat most often:
- Menu item descriptions for new arrivals, written to a fixed length.
- Order status messages that tell a customer when a driver is nearby without sharing unnecessary personal details.
- Replies to reviews, both positive and negative, that stay calm and do not argue.
- Weekly staff briefings that summarize which products sold, which were returned, and which questions came up most often.
- Training notes for new drivers on ID verification and handling a refused delivery.
Compliance Checks Before Anything Goes Live
Washington’s cannabis rules are set by the Washington State Liquor and Cannabis Board, and advertising, packaging, and delivery requirements can change. An AI prompt does not know which version of the rules is current, and it will not catch every issue. Every piece of generated copy should be reviewed by someone who knows the rules before it appears on your website, menu board, social feed, or text message list.
A simple review checklist helps:
- Does the copy make any health, wellness, or medical claim? If so, remove it.
- Would the language appeal to people under 21, such as cartoon references or youth slang?
- Does it include any required warnings or license information where your current rules require them?
- Are prices, potency figures, and product names copied exactly from your verified inventory system?
- Does it avoid implying that delivery is available outside your permitted service area?
Keep a record of which prompts were reviewed and by whom. If a regulator or a customer ever questions a piece of copy, being able to show a documented review process is far better than trying to reconstruct what happened.
Using AI for Customer Messages Without Losing the Personal Touch
Delivery customers often reach out with timing questions, address corrections, and requests to change an order. AI can draft fast, polite replies, but the tone matters. Ask the tool to write in short sentences, to avoid apologizing more than once, and to offer one clear next step. A reply to a customer whose order is delayed by weather on the South Hill should sound like a person who knows the roads, not like a corporate notice. To go deeper, explore The marketplace for AI prompts that actually work.
Always give the model real details when you can: the estimated window, the driver’s first name if you share it, and the reason for the delay in plain language. Vague placeholders produce vague messages. Specific inputs produce messages customers actually read.
Review Responses That Protect Your Reputation
Reviews are where a delivery business can lose ground quickly. A prompt for responding to a negative review should instruct the AI to acknowledge the specific problem, avoid defensiveness, avoid discussing the customer’s purchase history, and invite offline contact when appropriate. For positive reviews, the prompt should ask for a brief thank-you that mentions something specific from the review, so it does not read like a form letter.
Never let an AI reply publish automatically. Read every response, check it against your compliance checklist, and adjust anything that sounds too polished. Customers in Spokane tend to notice when a response feels canned.
Testing and Improving Your Prompts
A prompt is a draft, not a finished tool. Run each new prompt several times with different inputs and see whether the outputs stay consistent. If one version gives you a clean product description and another gives you a paragraph full of claims, find the difference and tighten the constraints. Save the version that works, note the date, and revisit it when rules or products change.
It also helps to have one person in charge of the prompt library. When everyone edits the same document without coordination, the library becomes cluttered and people stop trusting it. A monthly review of which prompts were used, which needed heavy editing, and which produced problems is enough to keep things healthy.
Getting Started This Week
You do not need to overhaul your operation to benefit from better prompts. Pick two tasks that take up the most time, write a prompt for each using the four elements described above, run them through your review checklist, and track how much editing the outputs need. If the drafts require little change after two weeks, expand to the next task. If they need heavy rewriting, the problem is usually the constraints, not the AI.
Good prompts will not replace your team’s knowledge of Spokane neighborhoods, your customers’ habits, or the rules you operate under. They are simply a way to get consistent, clear first drafts so your people can spend their time on the parts of the business that need a human judgment call. Start small, review carefully, and keep improving the library as your delivery service grows.
Leave a Reply