Many delivery teams in Quebec have started experimenting with AI assistants to draft replies, organize order notes and tidy up product descriptions, and the early results are often mixed. The difference between a useful answer and a generic one usually comes down to the instructions you give. If you are looking for professional chatgpt prompts that have already been written, tested and organized by task, a prompt marketplace can save time. Either way, the skill of writing a good prompt is worth learning, because it applies to every tool you will use.
Why most prompts disappoint
Most weak prompts share the same problem: they leave the model guessing. A request like “write a product description for our indica gummies” gives the assistant no audience, no tone, no legal boundaries and no format. The result is a vague paragraph that could describe any product from any company, and in a regulated category like cannabis, vague copy can also drift into claims you cannot make.
A prompt that works behaves more like a short brief to a new employee. It explains who the reader is, what the business does, what must be avoided and what the finished output should look like.
The four parts of a reliable prompt
1. Role and audience
Tell the assistant what job it is doing and who will read the result. For example: “You are drafting a reply for the customer support inbox of a licensed cannabis delivery service serving adults in the Montreal area. The reader is an adult customer asking about an order that is late.” That single sentence changes the tone, the vocabulary and the level of detail.
2. Context you would normally give a new hire
Paste the facts the assistant needs: your delivery windows, your cutoff times for same-day orders, your identification policy at the door, and the wording your team already uses. The assistant cannot know these things unless you provide them. Keep the context in a reusable block at the top of your prompt library so every prompt starts from the same verified facts.
3. Hard constraints
Constraints are where cannabis businesses should spend the most effort. List what the output must never include. Typical examples are medical or therapeutic claims, dosing suggestions, comparisons that imply health benefits, and any language aimed at minors. Quebec and federal rules on cannabis sales, packaging and promotion are detailed and can change, so your prompts should tell the model to flag anything that looks like a promotional claim, and your team should check the final wording against current official guidance. When in doubt, have a qualified lawyer review your standard messages.
4. A defined output format
Ask for exactly what you need: three subject-line options, a reply under 120 words, a bulleted checklist, or a table with columns for product, weight and sales note. Specific formats make outputs faster to review and easier to paste into your system.
Practical use cases for delivery operations
Used carefully, AI can remove repetitive writing from several parts of a delivery business. The key word is carefully, because every output still needs a person to approve it.
- Customer service drafts: Answers to questions about delivery windows, failed doorstep verification, substitutions or refunds. Staff edit and send.
- Order-status messages: Short, friendly updates that avoid promising exact arrival times you cannot guarantee.
- Internal checklists: Step lists for ID verification at handoff, sealed-package inspection and return procedures.
- Training material: Scenario questions for new drivers and support staff, with model answers that your compliance lead has reviewed.
- Website copy reviews: Asking the assistant to flag sentences that could be read as health claims or appeal to young people, so a human can rewrite them.
Building a prompt library your team will actually use
A prompt is only valuable if people can find it and trust it. Store prompts in one shared document or tool, label each one by task and date, and record who approved it. When a regulation, delivery zone or product line changes, update the context block once and every prompt that depends on it inherits the correction. To go deeper, explore The marketplace for AI prompts that actually work.
Test each prompt with at least five realistic examples before adopting it. Include awkward cases: an angry customer, a missing address, a request for medical advice. If the output fails any of them, tighten the constraints rather than hoping staff will catch every problem later.
Buying prompts versus writing your own
Purchased or shared prompts can be a good starting point, especially for general writing tasks such as summarizing a policy or rewriting a message in plainer language. They are less useful for anything that depends on your specific rules, delivery zones or brand voice. Treat marketplace prompts as drafts. Read them line by line, adjust the context to match your operation, and remove any instruction that conflicts with your compliance obligations.
A reasonable rule is to buy or borrow the structure and write the specifics yourself. The structure (role, context, constraints, format) is common to good prompts across industries. The specifics are what make the output accurate for your business.
Keeping humans in the loop
No prompt, however well written, replaces judgment. Cannabis is a regulated product, and customers may ask questions that touch on health, age or legality. Set a clear rule: anything involving health effects, dosing, age verification or a legal question is routed to a trained staff member, not answered by an assistant. Log the prompts you use and review them quarterly with whoever handles compliance.
A simple starting exercise
This week, pick one repetitive message your team writes every day. Write the prompt using the four parts above. Run it five times, score each output on accuracy, tone and compliance, and revise the prompt until the average score is high enough that a staff member only needs to make small edits. Then save it to your library and move on to the next task.
Good prompts are not clever tricks. They are clear instructions, written by people who know their business well enough to say exactly what they need. For a cannabis delivery service, that clarity is also a compliance tool, and it is one of the few areas where a little careful writing pays off every single day.
Leave a Reply