The useful result
Produce a friendly change-request reply that cites the agreement and offers options.
- Produce a friendly change-request reply that cites the agreement and offers options.
It starts with “one tiny extra thing”. Weeks later the project has doubled and the fee has not. Scope creep is rarely malicious; it is usually vagueness. The fix is a calm reply that points to what was agreed, says which part of the new request is outside it and offers clear options.
AI can draft that reply quickly, but only if it works from the real agreement. Anthropic’s guidance on reducing hallucinations suggests grounding answers in quotes from your documents, which keeps the draft from inventing terms. You supply the price and the timing; the assistant supplies the tone.
The examples below are illustrative exercises, not recorded results from any tool. Sources were checked on 24 October 2026; menus, limits and prices change, so confirm details on the linked pages.
Before you start
- Your agreement, quote or brief as text
- The client’s new request
- Your own price and timing for extra work
Review evidence
Documentation-based review · 2026-10-24
Primary documentation and official guidance checked on 24 October 2026. Examples, checks and exercises are original illustrations, not recorded tool tests. Cover artwork is a conceptual illustration.
Number the agreed scope
Paste the agreement’s deliverables, number of revisions, deadline and exclusions, and number each line S1, S2, S3. If the agreement is vague, note that too, because that is where creep grows. Do not rely on your memory of what was promised.
Check your result
The scope lines are numbered, and you have noted any vague or missing terms.
Classify the request and draft the reply
Ask the assistant to compare the new request with the numbered scope and label it inside scope, a clarification or outside scope. Then have it draft a short, warm reply citing the line numbers and offering options: do it as a paid change, swap it for something in the scope or schedule it for later. Leave price and timing as placeholders for you.
Compare the client’s new request with the numbered scope. First classify it as INSIDE, CLARIFICATION or OUTSIDE and cite the scope lines. Then draft a warm, short reply (under 130 words) that: thanks them, says what is covered, explains the new part is outside the agreed scope, and offers three options: paid change request, swap, or later phase. Use [PRICE] and [DATE] placeholders. Do not invent terms.
Scope:
[paste numbered scope]
New request:
[paste]Example / expected result
Illustrative example; not a recorded model result.
Scope S2 “two logo revisions”, request “a third round plus new colours”. Expected: classified OUTSIDE, with a [PRICE] option for a third round.
Check your result
The draft quotes scope lines, offers at least two options and contains placeholders for your price and date.
Fill the placeholders, read it aloud and keep a record
Add your real price and date, read the message aloud to check the tone and send it yourself. Save the thread with the agreement so the decision is documented. If the amount is large or the contract is complex, get proper advice.
Check your result
You have replaced every placeholder, the tone is friendly and firm, and the reply is saved with the agreement.
When it goes wrong
An AI draft can sound stiffer or friendlier than you mean, so adjust it. Do not let the assistant decide your prices. Contracts differ by country; this is not legal advice. Do not paste confidential client documents into tools you have not checked.
Make it your own
Write a four-line fictional agreement and a request that goes beyond it, then produce the reply and mark which scope line each sentence relies on.
Useful next resources
Sources and review
Last checked Oct 24, 2026. Features and account limits can change. Verify important results against your original material.
Did this help you complete your task?
Tell us what needs improving. No account or private output required.
Report incorrect information or a broken linkGenius Forges Editorial
Practical AI lessons with checked sources, labelled examples and human verification.
