Draft Quotes and Proposals in Minutes

A quote takes forty minutes and thirty-five of them are retyping. Your rates, your standard exclusions, the paragraph about access and parking, the payment terms. The five minutes that matter are the ones where you decide what the job actually involves.

The setup below moves the thirty-five minutes onto software and leaves the five with you. It uses one product, Claude, because a quote needs three things in the same place: your reference documents, a consistent voice, and a file at the end that a customer can open.

Load the project with the four documents a quote needs

Anthropic describes projects as "self-contained workspaces with their own chat histories and knowledge bases". The knowledge base is the point. Everything you would otherwise paste into every conversation goes in once and stays there.

Free accounts can create a maximum of five projects, per Anthropic's help page, which is plenty. Make one and call it Quotes. Then upload exactly four things:

1. Your rate card. Day rates, unit prices, call-out fees, minimum charges. Dated, so you can tell later which version a quote came from.
2. Two of your best past proposals. Real ones, sent and won, with customer names removed. These teach structure and phrasing better than any instruction you could write.
3. Your terms and exclusions. The list of what is never included. This is the document that saves you money, and it is the one most small businesses have never written down in one place.
4. Your scoping questions. The things you have to know before you can price anything. Access, timeline, who supplies materials, what happens if the spec changes.

Then set the project instructions. Anthropic's page describes these as a way to "define project instructions for each project to further tailor Claude's responses", giving the example of instructing Claude to answer "from the perspective of a specific role or industry". Paste something like this:

You draft quotes for a small [trade] business. Use only prices from the rate card in project knowledge. If a price is not in the rate card, write PRICE NEEDED in bold instead of estimating one. Always include the full exclusions list verbatim. Match the structure and tone of the two example proposals. Before drafting, check my scoping questions and ask me about anything I have not answered. Never state a delivery date I have not given you.

The "PRICE NEEDED" instruction is the important line. Left to itself a model will produce a plausible figure, because a quote with a gap in it looks unfinished and a quote with a number in it looks complete. You want the gap.

Switch on file creation before you ask for a document

By default you get text in a chat window. To get a file, the capability has to be on. Anthropic's page on creating and editing files with Claude gives the path: Free, Pro and Max users go to Settings > Capabilities and toggle Code execution and file creation on. Team and Enterprise administrators find the same toggle under Organization settings > Capabilities.

What comes out: "Claude can create Excel spreadsheets (.xlsx), PowerPoint presentations (.pptx), Word documents (.docx), and PDF files." The size ceiling is stated as 30MB per file for both uploads and downloads, and the page says code execution and file creation is available to all Claude users across Free, Pro, Max, Team and Enterprise plans.

So the ask at the end of a scoping conversation is a document, not a message:

Produce this as a .docx proposal. Structure: scope of work, itemised pricing table, exclusions, timeline, payment terms, acceptance line. Use only rate card prices. Leave the customer name as CUSTOMER for me to fill in.

Anthropic's own instruction on the same page is to "review and refine Claude's outputs to meet your exact requirements", which is not boilerplate when the output is a priced document you are about to sign your name to.

The security note on that page also applies directly to this workflow. Anthropic writes that "It is possible for a bad actor to inconspicuously add instructions via external files or websites that trick Claude" into unwanted behaviour. If a customer emails you a specification PDF and you drop it straight into your quotes project, you are uploading a document you did not write into a workspace holding your full rate card. Read what you upload.

Keep the price list in a workbook, not in the chat

Rate cards change and chats do not. If your pricing lives in a spreadsheet you already maintain, there is an add-in route that keeps the arithmetic where the arithmetic belongs.

Anthropic's page on using Claude for Excel describes it as "an add-in that integrates Claude into your Excel workflow", designed "for professionals who work extensively with spreadsheets, particularly in financial analysis and modeling". Availability is narrower than the chat product: "Claude for Excel is available to Pro, Max, Team, and Enterprise plans." Free accounts do not get it.

Installation, per that page: go to the Claude for Microsoft 365 listing on Microsoft Marketplace, click Get it now, then open Excel, activate the add-in and sign in with your Claude account. Administrators deploying it for a team go through the Microsoft 365 Admin Center at Settings > Org Settings > User owned apps and services, confirm that Let users access the Office Store is enabled, then add it under Settings > Integrated apps > Add-ins.

The feature that matters for quoting is that you can "ask questions about your workbook and get answers with cell-level citations". A cited cell reference is checkable in a way that a number in a chat reply is not. The page also lists what it will not handle: data tables, macros and VBA are unsupported file types.

And it carries two explicit warnings. Anthropic lists it as "Not recommended for: Final client deliverables without human review" and for "Audit-critical calculations without verification". It also says to "Only use Claude for Excel with trusted spreadsheets and not spreadsheets from external untrusted sources". A customer-supplied bill of quantities workbook is exactly an external untrusted source.

Two failure modes that cost real money

The total that does not add up. When a model writes a pricing table into a document, the total is text it produced, not a formula that calculated. It can be wrong by a line item and look completely normal. Check every total by hand, every time, or generate the pricing table as .xlsx with real formulas and paste the result in. This is the failure that turns a good week into an argument about whether your quote was binding.

The retired rate. You update your prices, you update the rate card on your desktop, and the file in project knowledge is still the old one. Nothing in the product will tell you. Put the effective date in the filename, make it the first line of the document, and add a standing instruction that Claude states the rate card date at the top of every draft. Then a stale card is visible in the output rather than buried in a workspace you set up in January.

There is also a limit nobody publishes. Anthropic's projects page says that when project knowledge approaches context limits, paid plans can "expand capacity by up to 10x while maintaining response quality", but there is no published figure for what that capacity is in words, pages or files. Four documents is comfortably inside any plausible reading of it. Forty years of past proposals is a plan built on an unstated number, and you will find the ceiling by hitting it.

One last thing about the five minutes you kept. The scoping questions document is the part of this setup that pays off longest, because it is the only piece that captures what you know and the software does not. Every time a job goes wrong for a reason nobody asked about, add the question you should have asked. The rate card is worth money. That file is worth more.