Customers often ask the same practical questions: What is included? How long does it take? What happens if something goes wrong? AI can help you organize clear answers from the policies you actually have.
The goal is a useful reference for a small business or freelance service, not a collection of pressure tactics. A customer’s concern may reveal that your offer is unsuitable or needs improvement.
What you will practise
Result: A draft customer answer tied to three fictional policy notes, with unresolved promises kept visible.
- Level: Beginner text practice.
- Time: About 10 minutes; an estimate, not a measured reader result.
- Device: Phone, tablet or computer; paper also works.
- Tools and cost: Use the complete on-page example for free, or an AI text chat you already use. No new account, upload, connected inbox, recording or paid feature is needed for this exercise.
Start with the fictional material. Applying the workflow to real work is separate and requires permission to use the source material.
Start with a small set of real questions
Choose five to ten recurring questions from notes you are allowed to use. Remove names, account details and unrelated personal information. You can begin with anonymized summaries instead of uploading entire call recordings or email archives.
Get appropriate permission before recording, transcribing or sharing customer conversations, and use tools approved for that work. A subscription to an AI service does not establish permission to upload somebody else’s information.
Connect each answer to an approved source
For every question, record the current answer, its source, the person responsible for approving it and the last review date. A source might be your published service description, signed scope or approved cancellation policy.
Leave unresolved points visible. If nobody has decided whether a revision is included, the answer should say that the owner needs to confirm it. AI should not fill the gap with a plausible policy.
Practise answering without inventing a policy
Try the task yourself, or select and copy the entire prompt below into a text chat you already use. All required source material is included. Do not add real customer, learner or other private information.
Use only the fictional notes S1–S3. Draft a brief answer to the customer question and place the supporting source ID beside each claim. Separate confirmed terms from points the owner must confirm. Do not offer discounts, extra revisions, a deadline or a satisfaction guarantee. Do not send anything.
S1: Maple Desk offers a one-page flyer layout for $120.
S2: The price includes one revision requested within seven calendar days after delivery.
S3: No delivery deadline or extra-revision price has been approved.
Customer question: Can I get two revisions and have the flyer tomorrow?
Help selecting, copying and pasting.
Illustrative answer — not a recorded AI response
Maple Desk’s one-page flyer layout costs $120. [S1] The price includes one revision requested within seven calendar days after delivery. [S2] I would need the owner to confirm whether an extra revision is available, its price and whether tomorrow’s delivery is possible. Those terms have not been approved. [S3]
Three checks before using the answer
- The answer preserves $120, one revision and the seven-calendar-day window after delivery.
- Two revisions and tomorrow’s delivery remain requests, not promises.
- Every policy claim has the correct source ID; no discount or guarantee is added.
Repair request if needed: “Restore the S1–S2 terms exactly. Move the second revision, its cost and tomorrow’s delivery under points requiring owner confirmation. Do not promise them.”
Explain the key check: Which part of the customer’s request is not covered by the confirmed price?
Finish when: the answer passes all three checks and you can explain the distinction above. Keep your checked draft; this exercise does not send, publish or approve anything.
Distinguish an answer from a sales promise
Suppose a fictional service includes one revision within seven days of delivery. An answer saying “We will keep working until you are completely happy” expands that offer into an open-ended promise. The approved limit should remain clear.
The FTC’s advertising guidance calls for truthful, evidence-based claims. Apply that discipline to promotional answers: do not turn “we provide a checklist” into “this will eliminate mistakes.”
If a customer says the service is too expensive, explain the actual scope and available options. Do not fabricate a limited-time discount or use guessed personal circumstances to pressure them. Sometimes the honest answer is that another option would fit better.
Test the library on awkward questions
Try a question involving an exception, an outdated price, an unclear deadline and something outside the service. Check that the draft asks for help where needed instead of producing a confident answer to everything.
A useful entry can include: the answer, the evidence, a follow-up question and the point at which a person must take over. Refund disputes, safety concerns and requests to make an exception should reach someone authorized to decide.
Keep it current and easy to use
Start with a shared document or your existing notes system. Retire outdated answers when a policy changes, and make the approved version obvious. Record questions that the library could not answer so the next update addresses real gaps.
Having an answer library does not make it safe to launch an unattended chatbot. Begin by using it to support replies that a person reviews. For a clear public explanation of the offer, use our sales-page exercise; for keeping track of agreed next steps, see customer follow-ups with AI.
Reviewed October 1, 2026. Prepared with AI assistance and checked against the sources linked in this article. Examples are illustrative. Read our editorial and information disclaimer.
Illustrative photo by Vitaly Gariev on Unsplash.
