If you run a small business, the value of Claude isn't "AI writes my emails." It's that the most important questions in your week get more of your attention, and the boring middle gets less. This page is the blueprint — every prompt copy-pasteable, the project template downloadable, the integration walked through end-to-end.
Block 60 minutes every Friday afternoon. (Or Sunday night. Pick a slot and don't let it move.) Use Claude on the single most important question of the week — not drafting emails. This is the practice that converts AI-curious owners into AI-leveraged owners. The output isn't a deliverable. The output is that you actually thought hard about the thing you've been avoiding.
The questions rotate, but the structure is the same. You sit down, open Claude inside your "My Business" Project, and pick one of the six questions below. You spend 45 minutes thinking with Claude and 15 minutes writing down the one decision you're going to act on next week.
If at the end of the hour you don't have one decision, something went wrong. Either the question wasn't right, or you let Claude drift into producing instead of provoking. The point is to make you work harder, not Claude.
"Walk me through every reason I should raise prices on [SKU/service] this quarter. Then walk me through every reason I shouldn't. Then tell me which side has the better argument."
"If I had to fire one customer this quarter, which would it be, based on the data I've shared, and what would the conversation sound like?"
"What hire am I postponing that I shouldn't be? Make the case for the hire, then make the case for solving the problem without hiring."
"My [vendor] renewal is coming up. Help me prepare to renegotiate — defensible ask, my BATNA, their BATNA, and the opening line."
"Look at the customers who've churned in the last 12 months. What pattern do I keep refusing to see? Be specific. Don't be diplomatic."
"If I had to spend $25k on the business in the next 30 days for the highest possible return on this business specifically, where would you spend it, and what's the case?"
Why this works. Most owners spend their week being asked questions. The Owner's Hour is the only hour where you ask the questions. Claude is a sparring partner that never gets tired, never has an agenda, and won't be in the room when you decide. That's a useful combination.
A Claude Project is a persistent context — everything in it is "remembered" across every conversation inside the project. Building one good Project is the single highest-leverage hour you'll spend in your first month with Claude. Every prompt below assumes this Project exists.
Download the template. Fill it in honestly — 30 to 45 minutes the first time. Paste the completed file into a new Claude Project. Open Claude → Projects → New Project → name it "My Business" → paste the content into custom instructions (or upload as a knowledge file).
Now every conversation inside this Project starts with Claude already knowing your business. You stop re-explaining yourself. You start getting answers calibrated to your actual situation, not generic advice for a generic business.
The "what I'm avoiding" section is the secret. Most owners build a project that's a polished brochure of the business — and end up with a Claude that sounds like a junior consultant. The version that earns its keep includes the hard stuff: the vendor you should have fired, the pricing you keep not raising, the review you keep not doing. Tell Claude these. Now when you ask it about something adjacent, it can call you on it.
Copy these into a notes file, into Claude as saved prompts, or paste them inline when you need them. They're written to be used inside your "My Business" Project so Claude already has the context — you just need to fill in the [bracketed] specifics.
You are my CFO. Using the data I've shared (P&L month-to-date, AR aging, AP aging, current cash), write me a Monday morning briefing with: 1. Cash position vs. last Monday (one number, one trend arrow) 2. Top 3 things I need to do this week to protect cash 3. One thing that's better than I might think 4. One thing that's worse than I might think 5. The single question I should sit with this week Length: under 250 words. Tone: direct, like a CFO who knows me well and isn't trying to spare my feelings.
Attached are this week's inputs: my calendar, this week's wins/losses, any context I've added. Draft my Friday all-hands update following our usual structure: - This week's wins (3–5, specific, name names where relevant) - This week's losses or learnings (1–2, honest) - What's coming next week (3 things) - The one number I want everyone to remember - One question for the team Tone: confident but not corporate — the way I actually talk. Length: under 400 words. If I'm being too soft on a loss, push back.
It's the end of Q[N]. Attached: the goals I set at the start of the quarter, actual results, last quarter's review, any added context.
Help me write my quarterly review:
1. Three things that went better than I expected — and why
2. Three things that went worse — being honest about bad luck vs. bad decisions
3. The one decision this quarter I'd make differently
4. The one decision I'd make again
5. Three goals for next quarter: ambitious, measurable, within my actual control
Push back if any of next quarter's goals aren't measurable or aren't within my control. I'd rather have three real goals than seven aspirational ones. I'm trying to decide whether to [DECISION].
Before I think about the answer, help me think about the question:
1. What's the actual decision I'm making? (Not what I think it is — what it really is.)
2. What's reversible vs. irreversible about this?
3. What information would change my answer? Do I have access to it?
4. What's the worst case if I'm wrong, and can I survive it?
5. What's the regret-minimization version of this question — what would I regret more, doing it or not doing it?
Don't recommend a choice yet. I want to think about the question first. I'm considering hiring [ROLE]. Help me think about this before I start writing the job description:
1. What problem am I actually trying to solve? Is hiring the right way to solve it?
2. What's the alternative to hiring? (Contractor, outsource, automate, do without, redistribute.)
3. If this hire works perfectly, what does my business look like 12 months from now that it doesn't today?
4. What's the full cost of getting this hire wrong? (Salary + onboarding + opportunity + cultural.)
5. What would I need to see in a candidate to feel confident this is right — three concrete signals?
Call me on it if I'm hiring to avoid a harder conversation about a different problem. We're renewing with [VENDOR] for [SERVICE]. I pay them [AMOUNT] per [PERIOD]. I've been with them for [DURATION]. Help me prepare: 1. What's a defensible ask (% reduction or added value)? 2. What's my BATNA — what does switching look like, realistically? Include the switching cost I'd rather not think about. 3. What's THEIR BATNA — what does losing me look like for them? 4. Three specific things I can ask for that don't cost them much but matter to me. 5. The opening line of the email. Be specific. "Negotiate hard" is not advice.
I need to write a difficult email. Here's the situation: [SITUATION] Here's what I actually want to say — including the parts I shouldn't write: [RAW THOUGHTS] Draft three versions: 1. Direct and professional — what I'd send if I fully trusted my judgment 2. Diplomatic — what I'd send if I'm trying to preserve the relationship 3. The version where I'm being too soft and need to be talked into version 1 After the drafts, tell me which one you'd actually send if this were your business, and why.
I need to have a hard conversation with [PERSON] about [TOPIC]. What I want from the conversation: [OUTCOME] What I'm worried about: [FEAR] What they're likely to say: [THEIR LIKELY POSITION] Help me prepare: 1. The one sentence that summarizes my position, said plainly 2. The three points I need to make, in order 3. The most likely pushback and my response to each 4. The line I won't cross — what I won't agree to 5. How the conversation ends if it goes well, and if it goes badly Don't soften this. I'd rather over-prepare for honesty than under-prepare and lose my nerve.
[CUSTOMER] is [N] days past due on invoice [#] for [AMOUNT]. We've sent [N] reminders. Our relationship is [describe: new/long-standing/strained]. Draft three escalating messages: 1. The friendly nudge (assumes they forgot) 2. The firm follow-up (assumes they're prioritizing other vendors) 3. The "this is becoming a problem" message (last one before I get the lawyer involved or stop work) For each, tell me when I should send it relative to today. And tell me which one to send first based on what I've told you about the relationship.
I have a meeting with [NAME] from [COMPANY] in [TIME]. The meeting is about [TOPIC]. Using public information (LinkedIn: [URL], company site: [URL], any recent news), give me: 1. Three things I should know about them before walking in 2. Two things I should know about their company that aren't obvious from the homepage 3. The most likely thing they actually want from this conversation 4. Three good questions for me to ask 5. One thing I should NOT say, given my position Length: under 400 words. I want to skim this in 90 seconds.
I'm reviewing the attached [CONTRACT TYPE]. I'm not a lawyer and this isn't a substitute for one — but before I send it to mine, I want to:
1. Identify the 3–5 clauses that are unusual or worth questioning compared to a typical contract of this type
2. List the obligations I'd be taking on, in plain English
3. List the obligations they'd be taking on, in plain English
4. Flag anything that limits my ability to exit the relationship
5. Flag liability, indemnification, or IP clauses I should specifically ask my lawyer about
Be conservative — I'd rather you flag something unnecessary than miss something important. End with the three questions I should ask my lawyer first. Attached is a proposal from [VENDOR] for [SERVICE]. My business does [WHAT]. We're considering this because [REASON]. Evaluate: 1. Does the proposed solution actually fit my stated problem? Or is it a generic deck dressed up? 2. What's the real cost over 12 months, including everything I'm likely to be charged extra for? 3. What are 3 specific questions I should ask before signing? 4. What's a realistic best-case ROI? Worst-case? 5. Is there a smaller version of this I should pilot first? If this looks like a bad fit for my business as you understand it, say so directly. I won't be offended.
How to actually save these. Three options. (1) Paste each into Claude as a "Custom prompt" inside your "My Business" Project — fastest to access. (2) Keep them in a single notes file and copy-paste when needed — most portable. (3) Star this page and come back. Anything that gets you to use them is the right answer; the worst version is the one where you read them once and never copy any.
One real integration is worth ten theoretical ones. This is the one to build first — high signal, low risk, and you'll feel the value the first Monday after it's live. Total setup: 30 to 45 minutes.
In Claude, open your "My Business" Project → Connect apps → QuickBooks → sign in with your QuickBooks Online credentials → grant read access. (If you don't use QuickBooks Online, the same pattern works with Xero or Wave — Claude has direct integrations for the major small-business accounting systems as of Anthropic's small business launch.)
Read access only. Claude can't post journal entries; it can read your books. That's all you need.
Copy Prompt 01 — Monday morning cash brief from the library above. Inside your "My Business" Project, save it as a Custom prompt named "Monday cash brief." Now it's one click away from every Monday morning.
Don't wait until Monday. Run the prompt now. Read the output. Look for two things: did Claude pull the right numbers (cross-check one or two against QuickBooks directly), and did the brief surface anything you didn't already know? If the answer to both is yes, you're done. If the answer to the first is no, tighten the prompt to name the specific reports you want it pulling from.
Cash this Monday: $84,210. Down $11,400 from last Monday — driven by two large vendor payments hitting the same week as a slow AR week.
Three things to do this week:
Better than you might think: gross margin month-to-date is 41%, up from 37% last month. The price increase on [SKU] is holding.
Worse than you might think: AR is creeping. Average days-to-pay is up to 38 from 31 in Q1. Worth a conversation with the bookkeeper this week.
Question to sit with: if Acme doesn't pay this month, what's the trigger that changes how we work with them?
Block 15 minutes on your calendar every Monday at 8am. Open the Project, run the saved prompt, read the brief, do the one or two things it tells you to do. Don't make this fancy — the value is in the rhythm, not the technology.
After four weeks, you'll have something most owners never have: a weekly cash review you actually do, that surfaces problems early enough to fix them. That's the integration.
What this isn't. It isn't an autonomous CFO. Claude doesn't decide anything. It pulls the numbers, applies a structured question to them, and hands you a 250-word document. You are still the CFO. The integration just means the prep work happens without you doing it. That's the right amount of automation for cash.
Once L1–L3 are real for you, L4 is where Claude operates on a schedule, surfaces things proactively, and reaches across systems. Here's the L4 sketch for an owner. Treat it as the destination, not the next step.
Every morning at 7am, before you're at your desk, Claude has already pulled from QuickBooks, your CRM (HubSpot, Pipedrive, whatever), your calendar, and your inbox. It cross-references against your "My Business" Project, the goals you set for the quarter, and the "what I'm avoiding" list. It DMs you a single Slack message:
Cash: $79,400. Flat vs. yesterday. AR action: Acme still 47 days out, escalation due today per our policy.
Pipeline: Two deals moved to "verbal yes" yesterday (combined $42k). One deal stalled at proposal stage for 21 days — that's your threshold; suggest a check-in email today.
Calendar today: 3 internal, 1 external (the call with [Vendor] at 2pm — your prep brief is in the thread). You blocked 4–5pm for the Owner's Hour. Q05 ("the honest churn") is queued — last touched 3 weeks ago.
Thing you're avoiding: the review with [team member]. It's been on the list for 27 days. Want me to draft the opening?
That's L4 for an owner. It's a real thing to build. And it's also a thing most owners shouldn't build until L1–L3 are habit, because the brief is only as good as the project context, the integrations, and the saved prompts feeding it.
If you're reading this and thinking I want that now — the honest answer is: spend a quarter at L1–L3 first. Then the L4 build is a weekend, not a six-month consulting engagement. Skip the foundation and the daily brief will be confident, polished, and wrong in ways you won't notice for months. When you're ready, that's a good moment for a 30-minute call.
Most small-business owners who try Claude get one good week, then drift back to old habits inside 30 days. Here's what kills it — named explicitly so you can see it coming.
If every conversation with Claude is "how do I do X" and you walk away with an answer, you're using it as a search engine and you'll plateau at a 1.2× bump.
The leverage is in the prompts above — decisions, drafts, conversations. The work that requires your judgment is the work where Claude saves you the most time, because it does the prep.
Owners who skip the "My Business" Project end up re-explaining their business in every conversation. Within two weeks, the friction outweighs the value and they quit.
The Project is the one-time tax that makes everything after it cheap. Pay it once.
The integrations and prompts produce time. If you don't have a ritual that uses that time on something high-leverage, you'll spend it on email and convince yourself Claude didn't help.
The hour is the productivity. Don't skip it because you're "too busy" — that's the week you most need it.
The trust-and-verify rule: anything that leaves your business — to a customer, the IRS, a contract counterparty — gets verified. Anything internal or in draft — trust and move.
Owners who verify nothing have an incident in their first month. Owners who verify everything quit in their second.
Reading about agents and trying to build one before the Project exists, before any prompts are saved, before a single integration is live. The agent will be confident and wrong, you'll lose faith, and the whole thing will end with a Slack message to your bookkeeper saying "this AI stuff is overhyped."
Walk up the rungs. The ladder exists for a reason.
The first three levels you can absolutely do yourself with this page. If you hit Level 4 — real automation, custom MCP, scheduled agents touching your data — and you're trying to figure it out at 11pm on a Wednesday, that's the moment to ask for help.
Not because you can't. Because your time is better spent on the parts of your business only you can do.
Download the template. Build the Project. Save the prompts. Wire the cash brief. Do the Owner's Hour. Come back in 90 days and the business you're running will be a different business — without any of the things consultants charge $50k to give you.