Claude for Quoting and RFQs: A Faster Quote Desk for PE-Backed Distributors and Manufacturers

1. The Short Answer

Claude for quoting and RFQs works as a drafting desk. Claude reads the RFQ email and every attachment, matches each line against your price book and past quotes, and drafts the quote and the cover email. The estimator or salesperson sets every price and approves every quote.

That split is the design. Most of a quote's hours go to assembly: opening attachments, finding parts, looking up last time's price, retyping. The judgment takes fewer hours and carries all of the risk, so it stays with your people.

Looking for consultants to build it? Have each firm draft from a batch of your past RFQs and compare the results with what your team actually quoted. Then buy the build at a fixed price, read-only connections first.

The tool-neutral maps of where AI pays are the playbooks for distributors and manufacturers. This guide is the Claude build.

Speed comes from the drafting. Margin still comes from the person who signs.

2. Why Quoting Is the First Place to Put Claude

Put Claude first where the work sits closest to revenue and is easiest to check. At most distributors and manufacturers, that is the quote desk.

Price aside, two things decide how much quoted work you win: how fast the quote arrives and how many RFQs you answer at all. Every busy desk has RFQs it never prices, because the estimators ran out of day.

The work in between is reading, matching and drafting, and each step produces something a person can check against a source. A wrong line in a draft sits in front of the estimator before any customer sees it.

Bain made the wider point in September 2026: one use case that reaches the P&L creates more value than dozens of pilots. The quote desk is the shortest route from Claude to the P&L. Here is the same desk, line by line:

Read across each row. A person stays in every one. The retyping is what leaves.

3. What Claude Does Well in a Quote

Five jobs, in the order an RFQ needs them:

• Reading the RFQ in whatever shape it arrives. An email thread, a PDF spec, a spreadsheet of line items, a scanned drawing. Claude reads text, tables and images, so the quantity in paragraph four of the spec lands on the list with everything else.
• Extracting the requirements. Part numbers, quantities, materials, tolerances, certifications and delivery dates, each with the page it came from, so the estimator checks a source instead of retyping it.
• Matching against what you already know. For a distributor, that means cross-referencing a customer's or competitor's part numbers to your SKUs and suggesting substitutes. For a manufacturer, it means finding the nearest past job and its cost sheet. Have Claude show why it picked each match.
• Drafting the quote and the cover email, in your template, with your terms and exclusions.
• Flagging what is missing. A drawing revision that differs from the RFQ, a missing quantity break, a material with no current cost. That list of questions for the customer is often the most useful thing on the page.

Claude works from what you give it. Feed it a stale price book and it drafts stale prices faster.

4. What Claude Must Not Do

Write these limits into the quote instructions, so they travel with every draft:

• Set a price on its own. Claude can show the list price, the customer's last price and your margin rule. The estimator or salesperson picks the number, because only they know the plant is booked through March or the account is about to place a blanket order.
• Promise a lead time. A date in a draft stays a placeholder until the planner or the system of record confirms it.
• Write to the ERP or CRM without approval. On Team and Enterprise, Owners set each connector's tools to Always allow, Needs approval or Blocked for the whole organization (Anthropic Help Center). Allow read-only tools and block write tools until the pilot has tested the review step.
• Send anything to a customer. Claude drafts the email, and a person reads it and sends it.

In Cowork, check two defaults listed on Anthropic's Team and Enterprise Cowork page. The option that lets members use "Automatically approve" mode is on by default, so turn it off before estimators use Cowork on live quotes. "Always allow" for connector tools is off by default; leave it off.

A quote is a promise with your company's name on it. Claude drafts it, and a person makes it.

5. The Data a Quote Desk Needs

Claude's drafts are only as current as the files behind them. Before anyone writes a prompt, find these five and decide which version is true:

• The price book. List and customer-specific prices, discount rules and surcharges. Usually in the ERP, with exceptions in someone's spreadsheet.
• Cost sheets. Rates and routings for a manufacturer, landed costs for a distributor. Usually split between the ERP and the estimators' own workbooks.
• Past quotes, with wins and losses. The most useful file, and often the worst kept: each quote with its outcome and, where anyone wrote it down, the reason.
• The catalog. Part numbers, cross-references, substitutes and discontinued items.
• Terms. Standard terms, freight rules and the exclusions legal approved, often in a document nobody has opened since the last acquisition.

Clean before you build. Pick one product line, export a year of its quotes, mark the outcomes and retire the duplicate price lists. Only your estimators know which spreadsheet is the real one.

Keep all of it on Claude Team or Enterprise. Anthropic's Commercial Terms say "Anthropic may not train models on Customer Content from Services." Free, Pro and Max run on consumer terms, where chats can train models unless the user opts out, and Pro and Max have no admin console. A price book is your margin in a file.

Anthropic's Cowork safety page adds: "Avoid granting access to local files with sensitive information, like financial documents." Cost sheets qualify, so keep them in a shared project rather than a laptop folder. Customer drawings raise their own questions, covered in the manufacturers playbook.

The model is the easy part to buy. The price book is the hard part to fix.

6. Building It on Claude

A quote desk on Claude has four parts, and only the connectors sometimes need building:

• A project per product line. A Claude project holds the files and instructions for one body of work: here, the line's price book, cost sheets, past quotes and terms, shared with the people who quote that line.
• A quote skill. The method, written once: what to extract, how to match, the margin rules, the template, the exclusions and the limits from section 4. On Team and Enterprise, skills can be shared organization-wide, so every estimator works from the same version.
• Excel, where the estimators already work. The cost buildup often lives in a workbook with years of logic in it. Keep it. Claude reads the workbook and fills in the inputs, and the formulas stay the estimator's.
• Read-only connectors. Where your ERP or CRM has a connector, or one can be built, Claude can look up stock, open orders and customer history while it drafts. An Owner enables each connector and starts it read-only, and each person still signs in individually. The MCP guide explains how connectors work.

Keep the method in the skill and the data in the project, and each can change without breaking the other.

Start with a project and the skill. Bring in Cowork, where Claude works through multi-step tasks and by default asks before acting, once the approval steps have run for a few weeks.

Automatic intake from a shared inbox, a screen inside the ERP or quoting at volume across branches calls for a custom build. The projects versus custom build guide shows where that line falls.

Most of the build is writing down what your best estimator already does.

7. The Approval Workflow

Write the workflow down before the pilot, as five steps with a named owner for each:

1. Draft. Claude reads the RFQ, matches the lines and drafts the quote, the cover email and the open questions. Owner: whoever runs intake.
2. Estimator review. Each requirement checked against its source page, weak matches fixed, judgment applied. Owner: the estimator.
3. Price check. The price confirmed against the margin rules and the customer's history, with large or low-margin quotes sent to the sales manager. Owner: named by value and margin.
4. Send. A person reads the cover note and sends the quote from their own email. Owner: the salesperson.
5. Log. The quote as sent, its value, the outcome and the reason for any loss, filed where the next draft can find them. Owner: the salesperson, checked weekly by the sales manager.

Step five is the one teams skip, and the one that improves the next draft. Every logged win and loss becomes a past quote Claude can match against.

Watch the reviews in the first week. If an estimator approves a long quote in seconds, the review has stopped, and the workflow needs a slower check, such as a sample of lines priced again by hand.

Approval steps protect you only while people read them.

8. Measuring It

Measure four numbers for 30 days before the pilot and 30 days after, on the same product line:

• Quote turnaround. Hours from RFQ received to quote sent, on every RFQ.
• RFQs answered. The share that got a quote at all. Declined and ignored RFQs count against it.
• Quotes per estimator per day. This shows whether drafting saves time or just moves it into review.
• Win rate on quoted work. Wins over quotes sent, by count and by value. It moves slowest and matters most.

If nobody records when an RFQ arrives, start today. The before number is the one you cannot rebuild later.

Thirty days will show turnaround and volume. Win rate needs a quarter, since buyers take weeks to decide.

Report all four to the operating partner on one page, with the RFQ log behind it. A number nobody can audit will not survive a board meeting.

If turnaround falls and win rate holds, you are answering more RFQs at the same quality.

9. Rolling It Across a Buy-and-Build

In a buy-and-build, every acquired company arrives with its own price book, quote template and idea of margin. You do not need to fix that before the next deal closes.

Split what travels from what stays local:

• What travels: the quote skill, with its extraction fields, approval steps and limits; the measurement page; the training.
• What stays local: each company's price book, cost sheets, past quotes and terms, in its own project, owned by its own estimators.

One skill, many price books. When the platform changes a rule, such as a minimum margin, the change goes into the skill once and reaches every desk.

On Team and Enterprise, an organization-wide skill keeps every estimator on the same version, and Owners can turn off user-created skills, so nobody quotes from a private copy with last year's margin rule. Separate portfolio companies should each keep their own Claude organization and data, with the same skill file installed in each.

Prove the four numbers at the platform company first, then bring each add-on onto the skill once its price book is clean. An add-on with three conflicting price lists produces three kinds of wrong quote, faster.

10. Who Can Build It and What It Costs

Four kinds of help can build a quote desk on Claude:

• Your own team. An estimator who uses Claude well, plus someone who owns the data. Fits one company with clean data and time to spare. Watch for a side project nobody finishes.
• Your ERP or quoting software vendor. Fits when the whole desk lives in that product. Watch for a tool that sees only its own system while your RFQs arrive by email.
• A specialist consultant that builds on Claude. Fits a company that wants the desk built on its own data and run by its own people. Watch for demos on someone else's data and hourly builds with no end date.
• A large systems integrator. Fits a platform replacing its systems anyway. Watch for the quote desk becoming one line in a multiyear program.

Whoever you hire should deliver, in writing:

1. A sizing of the gap on your own data: turnaround, coverage and where the estimators' hours go.
2. Drafts on a batch of your past RFQs, compared line by line with what your team quoted.
3. The quote skill in your template, owned by your company.
4. Read-only connectors first, and write access only after a tested approval step.
5. The approval workflow and the measurement page from sections 7 and 8.
6. Training for estimators and salespeople on their own quotes.
7. Documentation a new estimator can follow, and a named owner inside the company.

Buy it in two steps: a short paid diagnostic on your own RFQs, then a build at a fixed price scoped from what the diagnostic found.

Then the seats: on Anthropic's pricing page, a Team standard seat costs $25 a month, or $20 billed annually. Assume six people on the desk, an assumption for illustration, and the bill is 6 x $20 x 12 = $1,440 a year. Enterprise costs $20 per seat per month billed annually plus all usage at API rates, with a 20-seat minimum when bought self-serve.

The seats are the small line. The real work is the data, the skill and the habit of logging every outcome.

11. Where to Start

Start this week, with steps that need no software:

1. Pick one product line with steady RFQ volume and an estimator who wants to try this.
2. Start the RFQ log: date received, date quoted, value and outcome, for every RFQ on that line.
3. Pull a year of past quotes and mark each one won or lost, with a reason where anyone remembers it.
4. Settle the data rules: Team or Enterprise only, which files go in, and who approves a connector.
5. Test on past RFQs before any customer sees a draft.

Own several distributors or manufacturers? Start at the company with the busiest desk. The operating partner's Claude kit gives every company the same policy, settings and scorecard, and the manufacturers playbook shows what follows the quote desk in a plant.

The first draft Claude writes on your RFQs will need fixing. Every logged outcome gives the next one more to match against.