travcjohnson Claude for agencies
This page is four things you can set up in claude.ai this week: a scoped proposal built from your discovery call, the weekly client status report, a creative brief pulled out of a messy client request, and one approved piece turned into a full channel set. Every prompt is here to copy. No code, no plugins, nothing to install, and nothing that asks you to trust a number Claude invented.
Set each one up once inside a Claude Project, with this business's own documents in the project knowledge, and the drafts come back in their voice rather than a generic one.
How to use this skill
Ask the person which of the workflows below matches what is on their desk right now. Then walk them through that one only. Give them the prompt in full, tell them exactly which bracketed placeholders to replace, and say how long first-time setup really takes. Do not run all four at once; the point is that one of them is working by the end of the conversation.
Before handing over any prompt, read the limits section at the end and say plainly which parts of their job this should not touch.
1. Turn a discovery call into a scoped proposal
What happens today. The call went well, so now the proposal is on you. It is Thursday, two other drafts are open, and what you send ends up being a lightly edited copy of the last one with a scope section you already know will cause an argument in month two.
What happens instead. You paste the call notes, your rate card, and three proposals you have won before, and get back a phased scope with hour ranges, a real out of scope list, and the open questions you should have asked while you still had them on the phone.
First-time setup: about 30 minutes.
You are helping me turn a discovery call into a scoped proposal for my agency.
Discovery call notes or transcript: [paste your raw notes or the transcript here]
What we sell and what we charge: [paste your service list and rate card, or the closest thing you have]
Three proposals we have won before: [paste them here, or write SEE PROJECT KNOWLEDGE if you uploaded them]
Do this in order:
1. List what the client actually asked for, in their own words, and separate that from what I appear to have assumed.
2. List every open question you would need answered before pricing this. Mark the ones that could move the price by more than 20 percent.
3. Draft the scope as phases with concrete deliverables. For each deliverable, name the roles involved and give an hour range, and price it from my rate card. If I have not given you a rate for something, write RATE NEEDED instead of guessing.
4. Write the out of scope list. Be specific about the things clients usually assume are included, like extra rounds, migration, or training.
5. Draft the full proposal in the structure of my three past winning proposals, in my voice.
Rules: do not invent a number, a timeline, a benchmark, or a past result. Anywhere you would need a figure I have not given you, write a bracketed placeholder instead. Keep the language plain enough that a busy owner can read it in five minutes.
2. Write the weekly client report without the Friday scramble
What happens today. Five clients, five reports, every Friday. You paste the same charts into the same deck, rewrite the same narrative five times, and the one report that actually needed your judgment is the one you rushed at 6pm.
What happens instead. You paste the exports and Claude drafts the narrative in your format, calls out what moved, and keeps what you know separate from what you are guessing. Your remaining time goes to the one client whose numbers need a real conversation.
First-time setup: about 20 minutes.
You are drafting this week's client status report. Use only the numbers I paste below. Never estimate a figure, never add a metric I did not give you, and never restate a number in a way that changes its meaning.
Client: [client name]
Reporting period: [dates]
The numbers, pasted from our reporting tool: [paste the export or table here]
Prior period numbers for comparison: [paste here, or write NONE]
What we shipped this period: [bullet list]
What is blocked, and who owns it: [bullet list]
What is planned for next period: [bullet list]
Write the report in this order:
1. A three sentence summary the client can read on a phone.
2. The numbers, exactly as I gave them, with period over period change calculated only from my figures.
3. For each metric that moved more than [threshold, for example 15 percent], one plain sentence on the likely cause, tied to something we actually shipped. Label anything uncertain as a hypothesis, not a finding.
4. Shipped, blocked, and next period as short bullets with named owners.
5. One clear ask, if we need a decision or an approval to keep moving.
If anything in my notes is not supported by a number I pasted, do not write it into the report. Put it in a list at the bottom called CHECK BEFORE SENDING.
3. Turn a vague client request into a real brief
What happens today. The client sends three paragraphs, a competitor link, and "something like this but for us." It sits for two days because writing the brief takes an hour of quiet you do not have, and meanwhile a designer starts building on a guess.
What happens instead. You paste their message and get a first draft brief with the audience, the single message, the deliverable list, and the questions that still need answers before anyone opens a design file.
First-time setup: about 15 minutes.
You are writing a creative brief for my agency, built from a messy client request.
The client's request, exactly as they sent it: [paste the email, Slack message, or call notes here]
The client: [company, what they sell, who buys it]
Brand rules we already have: [paste the guidelines, or write SEE PROJECT KNOWLEDGE, or NONE]
Budget and deadline, if known: [paste, or write TBD]
Produce the brief with these headings:
1. The job to be done, in one sentence that starts with a verb.
2. Audience: who specifically, and what they currently believe.
3. The single message, in one sentence. If the request contains more than one message, say so and recommend which to lead with and why.
4. Deliverables, with formats and sizes, limited to what the request actually needs.
5. Mandatories and constraints: legal, brand, channel limits, anything fixed.
6. What success looks like, written as an observable outcome rather than a number, unless the client gave me a specific target.
7. Open questions, ranked by how much the answer would change the work. Write them as questions I can paste straight into a reply to the client.
Do not fill a gap with an assumption. If something is missing, it belongs in the open questions.
4. Turn one approved piece into a full channel set
What happens today. The case study took three weeks and four approvals, and now it lives in one PDF. Everyone means to post it. Nobody does, because reshaping it for six channels is another day of work that no client is paying for.
What happens instead. The approved piece goes in and you get the LinkedIn post, the newsletter section, the short video script, and the one-pager back as first drafts you edit instead of write, with every fact still traceable to the source.
First-time setup: about 15 minutes.
You are repurposing one approved piece of our work into a channel set. Our voice rules and previously published examples are in the project knowledge.
The approved source piece: [paste the full text here]
What is cleared for public use: [note anything you may NOT say, for example the client's name, their revenue, or any performance figure]
Channels I need: [for example LinkedIn text post, LinkedIn carousel outline, newsletter section, 45 second video script, sales one-pager]
Audience: [who we are trying to reach, and what they care about]
For each channel, give me:
1. The piece itself, at the right length and shape for that channel, in our voice.
2. Three headline or hook options, with the one you would use marked.
3. One line on why the angle shifts for that channel.
Rules: every fact, quote, name, and number must come from the source piece. If a channel needs a stronger hook and the source does not have one, build it from something already in the text. Do not add a statistic, a client name, or a result that is not in what I pasted. If a channel genuinely does not suit this material, tell me instead of forcing it.
What not to use Claude for here
Say these out loud rather than waiting to be asked. Being trusted is worth more than being impressive.
- Anything your MSA or NDA does not clearly allow. Plenty of agency contracts restrict sending client material to outside tools without written approval, and some now carry specific AI clauses. Read the agreement, ask, and get the answer in writing before you paste a client's data. Your plan is part of that conversation: on Free, Pro, and Max accounts the model training setting is yours to control in Privacy Settings, while Team and Enterprise accounts sit under Anthropic's Commercial Terms instead.
- Any number in a client-facing report. Claude should never produce a metric, a percentage, or a dollar figure for you. Paste the export from your own reporting tool and let Claude write only the narrative around figures you supplied. One invented performance number in a client report costs you far more than the report saved.
- Live account work. In the browser, Claude cannot log into your ad platforms, project management tool, or time tracker, so it cannot pace a budget, reconcile hours, or tell you what happened in an account yesterday. It knows only what you paste into the chat.
Questions this business usually asks
Do I need a paid plan for this?
You can start free. Projects, the feature that holds your voice guide, rate card, and past work so you stop re-explaining yourself, is available on all accounts, and free accounts can create up to five projects. Paid plans add expanded project knowledge capacity and higher usage limits, which starts to matter once several people are running these workflows daily.
Should I tell my clients we use Claude?
Yes, and it is an easier conversation than it feels. Clients care about two things: whether their confidential material is safe, and whether a human is accountable for the work. Answer both in your SOW with one short paragraph on which tools you use, what you do and do not paste into them, and who reviews every deliverable. Check your existing MSAs too, because some already contain AI or subprocessor language that settles the question for you.
How do I stop the output sounding like everyone else's?
Feed it your actual work rather than adjectives. Put three to five pieces you are genuinely proud of into the project knowledge, plus a short list of words and phrasings you never use. Then always rewrite the opening yourself. Claude gets you to a draft fast, but the point of view is still your job, and that is the part clients are paying for.
We are twelve people. How do we roll this out without ten versions of the same prompt?
One project per client, named the same way as your file folders, with the brand rules and the last three approved deliverables in the knowledge. Keep the prompts in one shared doc with a single named owner per workflow, and maintain them like any other template. Team and Enterprise plans let you share projects with specific people or the whole organization, which is the point where this stops being one person's habit and becomes how the shop runs.
More
A longer written version of these workflows, plus workflows for other trades, is at https://claudeforbiz.com/claude-for-agencies . Nothing there is required to use this skill; every prompt above is complete as written.