đ Works best with your inbox. Connect Gmail and Jeremy AI reads the client's request emails and drafts each reply. It never sends.
Paste your stated scope and the requests a client sent. Get one file back: a scope table with a status for every request, a reply draft for each one, a not-provided list and a check table. Nothing is sent.
You'll see the SAMPLE scope and requests pasted in, the scope check built in about 4 seconds on screen (the real wait was about 71 seconds), the chat summary, the finished file, the scope table with a status for every request, the reply drafts with their greeting and sign-off placeholders, the not-provided list and the check table scrolling down to the last row.
What this use case does and who it's for
Scope creep rarely arrives as one big ask. It comes as "Can you just add a fourth campaign?" and "Can we hold the call for 45 minutes?" Each one feels small. Each one needs a calm, clear answer. And writing those answers one by one is slow.
Here's the shortcut. Paste your stated scope and the client's requests. Jeremy AI checks every request against the scope and drafts a reply for each. The drafts are kind, short and never promise extra work. They don't state prices or deadlines. Four parts:
A scope table. Four columns: request, status, scope item and reply type. One row per request, in the order you gave them.
Reply drafts. One per request, 45 words or fewer. Each starts with a client-name placeholder and ends with a sign-off placeholder on its own last line.
A not-provided list. Everything the replies would need that you didn't give, like the client name, account lead, scope change process, report delivery day and contract terms.
A check table. Three columns: item, source and status. It shows the request wording and scope items were copied exactly.
The status rules are simple. In scope means the request matches a stated scope item exactly. Out of scope means it asks for more than a stated item, or something not listed. Needs detail means it matches an item but asks for a detail you didn't state, so the reply asks one question.
It's for agency owners, freelancers, consultants and account managers. They're drafts. You read them, fill in the placeholders and send them yourself. Jeremy AI doesn't send anything.
What you need before you start
A Jeremy AI account. Nothing to connect to try the sample.
Your stated scope as a short list. The sample is made up: SAMPLE Lumen Growth, a paid-ads agency for gym owners, with four scope items: 3 ad campaigns, a weekly performance report, 2 rounds of creative edits and a monthly strategy call of 30 minutes.
The client's requests, each with an id. The sample has five (Q-01 to Q-05), like "Can you add a fourth campaign?" and "Can you build us a landing page?"
Honest gaps. The sample says client name, account lead, scope change process, report delivery day and contract terms are "not provided."
The hard rules. The sample uses HARD RULES: copy request wording and scope items exactly, never state a price, fee, discount, bonus, deadline or timeline, never promise a result or extra work, never invent a scope item or policy, never calculate a number, and never address you by name.
How to set it up
You use the default Jeremy AI worker. There's nothing to build.
Open Jeremy AI. From the dashboard, type in the main box. Check the worker selector says Jeremy AI.
Paste the prompt below. Swap your own scope, requests and gaps in where the sample ones are.
Click Send message.
Wait. The walkthrough speeds this part up. In real time it took about 70 seconds.
When it says the file is saved, read the chat summary. It tells you which requests are in or out of scope.
Open the file card to see the finished scope file.
Read the scope table, then the reply drafts, the not-provided list and the check table.
The prompt
This is the exact prompt from the walkthrough. The agency, scope and every request are SAMPLE and made up. Replace them with yours.
Act as my "Scope Check" worker for SAMPLE Lumen Growth, a paid-ads agency for gym owners (all data below is made up). Check the SAMPLE client requests below against the stated scope and draft a reply for each. Create ONE file named scope-check.md using the file tool (reply with a short summary only) with: 1) a SCOPE TABLE with exactly 4 columns (request, status, scope item, reply type), one row per request in the order given, 2) REPLY DRAFTS, one per request in the order given, each of 45 words or fewer, 3) a NOT PROVIDED list, 4) a CHECK table with exactly 3 columns (item, source, status). STATUS RULES: in scope = the request matches a stated scope item exactly (scope item cell copies that item); out of scope = the request asks for something beyond a stated scope item or not listed (scope item cell says "none"); needs detail = the request matches a scope item but asks for a detail not stated (copy the item, and the reply asks one question). REPLY RULES: in scope = confirm it is included; out of scope = say kindly it is outside the current scope and point to "[scope change process: not provided]"; needs detail = confirm the item and ask one question about the missing detail. Every draft starts with "Hi [client name: not provided]," and ends with the sign-off "[account lead: not provided]" on its own last line, never inside a sentence. REPLY TYPE CELLS: write exactly "In scope: confirm it is included" for in scope, exactly "Out of scope: point to scope change process" for out of scope, and exactly "Needs detail: confirm item, ask one question" for needs detail; never use the word decline. Every sentence in every draft starts with a capital letter, including the first sentence after the greeting. HARD RULES: copy request wording and scope items exactly; never state a price, fee, discount, bonus, deadline or timeline; never promise a result or extra work; never invent a scope item, policy or detail not stated; never calculate a number; write "not provided"; every table has 4 columns or fewer and cells of 12 words or fewer; never address me by name anywhere in the chat reply or the file. Drafts only, nothing is sent.
STATED SCOPE (SAMPLE):
- 3 ad campaigns
- a weekly performance report
- 2 rounds of creative edits
- a monthly strategy call of 30 minutes
CLIENT REQUESTS (SAMPLE):
- Q-01: "Can you add a fourth campaign?"
- Q-02: "Please send the weekly performance report on Fridays."
- Q-03: "Can we do a third round of creative edits?"
- Q-04: "Can we hold the monthly strategy call for 45 minutes?"
- Q-05: "Can you build us a landing page?"
Not provided: client name, account lead, scope change process, report delivery day, contract terms.
Prompting tips and how to adapt it
Write your scope as short items. "3 ad campaigns" is easy to match. A long paragraph isn't.
Give each request an id and keep the client's words. The table copies them, so you can match every row to the message.
Define the three statuses. In scope, out of scope and needs detail. Jeremy AI follows the rules you write.
Fix the reply type wording. The prompt spells out the exact reply type cells. That keeps the table consistent every time.
Point to your process with a placeholder. The drafts say "[scope change process: not provided]." Swap in your real process before you send.
Ban numbers you didn't give. The prompt says never to state a price, fee or timeline. You decide those yourself.
Cap the length. 45 words or fewer per draft. Short replies read as calm.
Adapt it. Use it for design retainers, coaching packages, web projects or support plans. Swap the scope list and keep tables at 4 columns or fewer.
The result to expect
One Markdown file named scope-check.md, titled "Scope Check: SAMPLE Lumen Growth." With the sample data you should see:
A 4-column scope table with five rows.
Q-01 "Can you add a fourth campaign?" is out of scope. Scope item: none. Reply type: out of scope, point to scope change process.
Q-02 "Please send the weekly performance report on Fridays." needs detail. Scope item: a weekly performance report. The delivery day was never stated, so the reply asks one question.
Q-03 "Can we do a third round of creative edits?" is out of scope.
Q-04 "Can we hold the monthly strategy call for 45 minutes?" is out of scope.
Q-05 "Can you build us a landing page?" is out of scope.
Five reply drafts. Each starts with "Hi [client name: not provided]," and ends with "[account lead: not provided]" on its own last line. The out-of-scope drafts say kindly that the request is outside the current scope and point to "[scope change process: not provided]." The Q-02 draft confirms the report and asks which day to deliver it.
A not-provided list. Client name, account lead, scope change process, report delivery day and contract terms.
A check table. Each item with a source and a status of Pass, such as request wording copied exactly, scope items copied exactly, no price or timeline stated and drafts of 45 words or fewer.
The chat reply is a short summary. It says one request needs detail, four are beyond scope, and nothing was sent.
These are sample results from made-up requests. Wording changes from run to run. Read every draft before you send it.
Common mistakes and fixes
A draft says "decline" or sounds cold. Rewrite it. Remind it: "Say kindly it is outside the current scope."
A draft quotes a price or a timeline. Delete it. Remind it: "Never state a price, fee or deadline."
A draft promises extra work. Delete it. Remind it: "Never promise extra work."
A request is marked in scope when it isn't. Check it against your scope list. A request is in scope only if it matches an item exactly.
A draft invents a policy. Replace it with the placeholder. Remind it: "Never invent a policy or detail."
The sign-off sits inside a sentence. Tell it: "The sign-off goes on its own last line."
You send without filling the placeholders. Don't. Add the client name, your name and your real scope change process first.