Can AI speed up quoting when you quote from spreadsheets?

The problem, briefly

Excel quoting breaks down in familiar ways. The price list lives in one file and the spec sheets in another. Someone copies numbers across by hand. Customer-specific prices sit in a head or a sticky note. A quote goes out late, and a slow reply can cost you the job. I plan to cover the full choice between quote software and Excel in a separate guide. Here the question is narrower: what can AI do about the slowness?

The slow part is rarely the arithmetic. It is the reading and matching. A customer emails a list in their own words, and a person works out which of your products each line means. That reading and matching is work a language model does well as a first pass.

Where AI genuinely helps

• Reading a messy request. Customers send a PDF, three lines of email text or a photo of a list. AI can pull out each item, the quantity and any deadline into a clean table. Claude's help center says it reads both text and visuals in PDFs of up to 100 pages, and text only in PDFs of 101 to 1,000 pages.
• Matching lines to your price list. Paste your price list and the extracted request, and ask for the closest product for each line. It can flag lines that match more than one product, or none.
• Reading spec sheets. A model can check whether the product you matched meets the size, rating or material the customer asked for, and point to the line in the spec sheet that supports it.
• Drafting the quote text. Once the lines and prices are set, AI can write the covering email and the terms in your usual wording.
• Flagging odd quotes. Ask it to check a finished quote against last time's: a price 30 percent different, a quantity far off normal, a missing item. It is a second pair of eyes, not a rule engine.

What it should not do alone: set a price, apply a contract discount it was not shown, or send anything to a customer. Those stay with a person. The arithmetic is also better done by your spreadsheet formulas than by the AI, so use AI to match and read, and let Excel add up.

Is there a real example of a business doing this?

There is no independent, published case study with named results for AI speeding up quoting from spreadsheets. What exists comes from vendors, and the figures in it are self-reported, so they are not repeated here.

Here is how one vendor says it works, because it matches the steps above. Airdev, a firm that builds custom AI order entry for distributors, describes its approach for companies running Datacor, an ERP system (business software for orders, stock and accounts) used by chemical distributors. The system picks up an emailed request, matches each requested product, prices it from the account's actual price book, contract pricing and quantity breaks, and stages a draft quote for a person to review. Airdev says nothing reaches a customer unreviewed. That is a description of a process, not a result, but it shows the shape that works: AI drafts, a person approves. A post on the blog of NAED, the electrical distributors' trade association, written around a software vendor's own AI product (Ziffity), makes the same point about the mess of formats requests arrive in: emails, PDFs, spreadsheets and scans.

How to do this yourself with Claude

Claude is an AI assistant from Anthropic that you can use in a browser. This works best for a business with a price list of up to a few hundred lines.

1. Prepare the price list. One row per product, with a code, a name, the unit and the price. Remove cost and margin columns before pasting if you do not want them shared. Check your own confidentiality rules first. You can also upload the file instead of pasting: Claude's help center lists CSV among supported file types, and says Excel (XLSX) files need code execution and file creation switched on in your account. Another Claude help page says that setting is on by default on the Free, Pro, Max and Ultra plans, and that on Team and Enterprise plans an organization owner can turn it off, so if XLSX uploads fail at work, ask your account owner. For PDF price lists, Claude's developer documentation says password-protected PDFs are not supported, and that dense PDFs with small fonts, complex tables or heavy graphics can fill the context window before the page limit, in which case you should split the document into sections.
2. Paste a real request. Use one the customer actually sent, with their name removed.
3. Ask for a draft. A prompt that works: "Below is our price list, then a customer request. For each line in the request, find the matching product in the price list. Output a table with: customer wording, matched product code, unit, quantity, unit price. If a line could match more than one product or none, do not guess: list it separately with the options. Do not add up totals." The instruction to list unsure lines instead of guessing matters: Anthropic's guidance on reducing wrong answers says to give Claude explicit permission to say it is unsure.
4. Check the unsure lines. These are where your knowledge matters. Fix them yourself.
5. Add up in Excel. Paste the table into your quote template and let the formulas calculate totals and tax. Models can slip on sums, so keep the math in the spreadsheet.
6. Time it. Quote five real requests the old way and five this way, and compare minutes per quote and mistakes found.

If the results are good, you may not need new software yet. If quoting still hurts, see my guides to a quote generator and instant quote software for the ready-made options, and my Excel to web app guide for when the spreadsheet itself has outgrown its job.

Where a person still has to check

• Prices. Special deals, contract rates and quantity breaks need to be in what you paste. If they live in someone's head, the AI cannot see them.
• Fit. Two products that look alike may not be interchangeable. A person who knows the work signs off.
• Totals and tax. Check them in the spreadsheet.
• Sending. A person reads the quote before the customer does.

AI also does not remove the cause of slow quotes if the price list itself is out of date. Fix the list first. A model that matches against old prices produces old prices faster.

How I'd approach this

I would try the manual version for two weeks on real requests before spending anything. It shows you quickly how much of your quoting is reading and matching, which AI helps with, and how much is judgment, which it does not. To choose the first thing to try rather than guessing, the AI Strategy Sprint is a paid 3-day first step where we find where AI could help your business, and you get the exact steps written down. It starts with a free 15-minute AI brainstorm call.

If it works and you do it daily, the next step is a small tool that does the same steps automatically: it reads the incoming request, drafts the quote from your price list and sends it to you for approval, with the quote history stored somewhere searchable. Fixed price, agreed in writing first. If a free habit solves it, I will tell you.