agentclaw

workflow: quoting

The first decent quote usually wins. Yours is still in someone's inbox

Quoting looks free because nobody invoices for it. Look closer: a request lands, waits for the one person who knows the pricing, gets built by hand from a price list that may or may not be current, and goes out two days later — after the buyer already has other numbers in hand. We install a system that reads the request, matches it to your catalog, prices it by your actual rules, and hands your team a ready-to-send draft in minutes. Your people approve. Software does the assembly.

Get your free AI audit

the manual version

Where the time actually goes

Follow one quote from 'can you price this?' to 'sent' and watch what the work really is. Someone reads a rambling email with a spreadsheet attached, or a PDF spec, or a voicemail transcription, and translates it into line items and quantities. Then the lookup chain starts: the catalog price, this customer's negotiated tier, the quantity break, the current freight rate, the lead time from whoever owns inventory. Each answer lives in a different place, and at least one lives in a specific person's head. The quote gets typed into a template, checked against a price list that was updated sometime (nobody is sure when), and routed for a gut-check before it goes out.

Run your own numbers. If a quote takes 45 minutes of combined effort and you send 60 a month, that is a workweek of skilled time spent on lookups and formatting. Then add the costs no timesheet shows: the request that sat two days in a full inbox, the discount someone improvised to move faster, and the sent quotes nobody ever followed up on.

Almost none of this is judgment. It's reading, matching, and arithmetic against rules you already wrote down — or meant to. That is exactly the work software should be doing, with your people handling only the calls that genuinely need a human.

  • The decoding: turning a messy email and an attached PDF into actual line items and quantities
  • The lookup chain: catalog price, customer tier, quantity breaks, freight — each in a different system
  • The bottleneck: one person who 'knows the pricing', and a queue forming behind them
  • The stale-price tax: quotes built from last quarter's list, discovered at invoice time
agentclaw · workflow run

$ claw run invoice-intake

→ 47 documents queued

→ extracted · matched · posted

✓ done in 3m 12s · 0 exceptions escalated

the automated version

From request to priced draft, by your rules

Three stages, wired into your inbox, catalog, and CRM. The draft is ready while the buyer is still waiting on your competitors.

  1. 01

    Read the request

    An agent watches wherever requests arrive (shared inbox, web form, portal) and reads the whole thing: the email body, the attached spreadsheet or spec PDF, the thread history. It extracts line items, quantities, sizes, delivery location, and deadline, then matches each item to your actual catalog and identifies the customer and their account terms. When something load-bearing is ambiguous, like a part number that matches two SKUs or a quantity that doesn't parse, it asks the requester or your team a pointed question instead of guessing.

  2. 02

    Price by your actual rules

    Prices never come from the model's imagination. Your price list, customer-specific tiers, quantity breaks, margin floors, freight logic, and validity windows are encoded as deterministic rules, applied the same way every time. The agent looks prices up and does arithmetic; it cannot invent a number. The output is a complete draft quote in your template or quoting system: line items, unit prices, totals, lead times, terms, expiry date.

  3. 03

    Flag, approve, follow up

    Anything outside the rules gets flagged, not filled in: margin below your floor, an item you don't stock, a customer on credit hold, a quantity past your discount threshold. Each flag routes to the right approver with the reason attached. A person reviews and sends — or, for clean quotes inside thresholds you set, sending can be automatic. Every quote is logged to the CRM, and unanswered ones get a scheduled follow-up instead of silence.

Signs it's time to automate this

You don't need all six. Two or three usually means the math already works.

  • Quote requests routinely wait a day or more while competitors answer the same afternoon
  • The same product goes out at different prices depending on who built the quote
  • Quoting stops when one specific person is out, because the pricing lives in their head
  • You find unapproved discounts and below-margin quotes after the fact, not before they're sent
  • Quotes go out with wrong SKUs or prices from an outdated list, and get caught at invoice time
  • Nobody can say how many quotes went out last month or how many turned into orders

Straight answers

Our requests are a mess — half emails, half PDFs, some phone calls. Can it handle that?+

That mess is most of the job, and it's the part modern AI is genuinely good at: reading unstructured text and pulling out items, quantities, and constraints. The agent parses email bodies, attachments, and call transcriptions, then matches what it finds against your real catalog. The honest caveat: matching is only as good as your product data. If your catalog has duplicate SKUs or three names for the same item, the first weeks of the engagement include cleaning that up, because ambiguous matches get escalated to your team rather than guessed.

What stops it from quoting the wrong price or giving away margin?+

The model never sets a price. Your price list, tiers, breaks, and margin floors are encoded as deterministic logic — lookups and arithmetic, the same way a spreadsheet works. The model's job is reading the request and writing the words around the numbers. When a request falls outside the rules, the draft ships with that line flagged and routed to whoever owns pricing. A manual quoting desk leaks margin through improvisation; a rules-based one can't improvise.

Can I do this myself?+

A light version, yes. With off-the-shelf AI tools you can paste a request into a chat window alongside your price list and get a decent draft, and our free skills library shows you how to build repeatable workflows like that for extraction, drafting, and follow-up. Where DIY runs out of road is trust and volume: pricing rules the model cannot override, catalog matching across thousands of SKUs, flags that route to the right approver, and CRM logging that happens every time. Pasting into a chat gets you a draft. It doesn't get you a quoting desk your whole team can hand requests to.

What does it cost?+

Engagements start at $5,000/month, which covers building the workflow and running it: keeping the catalog matching sharp, updating the pricing rules when your lists change, and fixing things when your tools do. Do your own math before booking. Count the hours your team spends building quotes each month, price them at loaded cost, and add an honest estimate of what slow turnaround and quiet margin leaks cost you. If that total sits well under the fee, you don't need us yet — start with the free resources and come back when volume grows.

Keep going

AI agents for wholesale distribution

Thousands of SKUs, tiered customer pricing, and quantity breaks: the setting where manual quoting hurts most.

AI agents for construction and trades

Where the job quote is the sales process, and the slow bid is the lost bid.

Workflow automation

How we wire agents into the inbox, catalog, and CRM you already run on.

ROI calculator

Put your own quote volume and hours in, and see what the manual version costs you.

Find out what your quote turnaround actually costs

The free AI opportunity audit maps how requests become quotes in your shop today, and tells you whether this is the workflow to automate first — or something else is.

Book your free audit

We take on companies ready to invest $5,000+/month. Not there yet? Our free resources are genuinely free.