Skip to content
All insights

Workflow automation

From an RFQ in the inbox to a quote worth sending.

The request arrives as an email with a specification attached. Somewhere between that email and a quote your team is willing to send, an experienced person spends forty minutes moving between a PDF, the catalog and the price list. That is the process worth examining.

Published November 18, 20257 min read

What this covers

  • Response time is usually the commercial problem, and it is caused by queueing rather than by the quote itself being hard.
  • Matching customer part numbers to your catalog is the step that actually decides whether automation works.
  • Pricing must be read live from the system of record, never from a copy, or the automation creates a commercial exposure.
  • Engineering-priced and non-standard lines stay with people, and a good deployment makes that split explicit rather than pretending otherwise.

There is a specific kind of loss that never appears in a pipeline report. A buyer sends the same request to three suppliers on a Tuesday morning. Two reply by Wednesday. The third replies on Friday, with a better price, and loses. Nobody records it as a pricing failure, because it was not one.

Quoting is the highest-leverage workflow in most B2B distributors and manufacturers for exactly that reason: the work is not difficult, it is just queued behind other work, and the queue is where the margin goes.

What actually happens between the email and the quote

Broken into its real steps, the work looks like this. Almost none of it is commercial judgment.

  1. 01Read the request

    Open the email, open the attachment, work out what is being asked for. The attachment might be a specification sheet, a bill of materials, a scan of a printed form or a spreadsheet with its own column headings.

  2. 02Identify the customer

    Find the account behind the sender. Pick up the contract, payment terms, pricing tier and anything on hold. If the sender is new, decide what terms apply.

  3. 03Match the products

    Convert what they asked for into what you sell. Customer part numbers, legacy codes, free-text descriptions and other people's units of measure all have to resolve to catalog items.

  4. 04Apply pricing

    Read what this account is entitled to pay: contract price, volume break, active discount. Not the list price, and not last quarter's copy of the price list.

  5. 05Check availability

    Confirm the quantities can be supplied by the date requested, from stock or inbound supply, and flag the lines that cannot.

  6. 06Assemble the quote

    Put it into the template, with terms, validity and the right delivery assumption.

  7. 07Route and follow up

    Send it if it is within policy, get approval if it is not, and chase it if nothing comes back.

Steps one through six are lookup, matching and rule application. An agent does all of them. Step seven is partly judgment. The commercial decision, which is what price to fight for on a deal that matters, was never in this list at all, and should not be.

Matching is where it is won or lost

This is the step that decides whether a quoting deployment works, and it is not a language problem.

A long-standing customer orders under part numbers that came from a supplier you stopped carrying in 2019, because their purchasing template has never been updated. Another sends descriptions with no code at all. A third uses your codes but drops the hyphens. A fourth asks for a quantity in feet when you sell in meters. All four are routine, and all four require knowing your catalog and that account's history rather than reading English well.

Reading the RFQ is the part that looks hard. Knowing which of your 40,000 items they meant is the part that is hard.

An agent resolves these by combining the text with the catalog, the account's order history and the specification attached. Where the evidence supports one item, it proceeds. Where it supports two, it stops on that line, shows both candidates and says why each was plausible. The other eleven lines are not held up while a person answers. Intelligent document processing goes further into why per-line confidence matters more than an overall accuracy score.

Pricing has to be live

This is the single technical requirement worth refusing to compromise on. Contract pricing, volume breaks and active discounts are read from the system of record at the moment the quote is prepared.

A copied price list ages, and it ages invisibly. A quote issued at a superseded price is either a margin loss you absorb or a retraction you explain to a customer, and an automation that produces them faster than a person did has made the problem worse rather than better. If live pricing is not reachable, that is the first thing a deployment fixes, before anything is automated. What an agent needs from your ERP covers what that access actually looks like.

What stays with your team

Being precise about this is not a disclaimer, it is the design. A quoting agent that pretends to handle everything ends up either wrong or held up on every case.

CaseHandled byWhy
Catalog items, contract-pricedAgent, end to endEvery value is a lookup or a rule
A line the agent could not match confidentlyPerson, on that lineThe agent shows its candidates; the rest of the quote proceeds
Engineering-priced or made-to-orderPerson, with preparationThe price depends on judgment the agent does not have; it assembles everything else
A strategic deal where price is a decisionPersonCommercial judgment was never the automation's job
A price below the floorPerson, by ruleValue threshold, regardless of how confident the agent is

The pattern in rows two and three is worth naming. The agent does not hand back an untouched request. It hands back a prepared one: customer identified, ten of twelve lines matched and priced, stock checked, with one specific question outstanding. That changes the human step from forty minutes to two.

What to measure

Four numbers tell you whether a quoting deployment is working, and only one of them is the obvious one.

  1. 01Time from request received to quote sent, measured from the email timestamp rather than from when someone opened it. This is the number the customer experiences.
  2. 02Share of quotes prepared without a person touching them, which is the number that converts into hours.
  3. 03Share of held cases that were held for a good reason. A high hold rate with routine approvals means a rule is too tight.
  4. 04Quote volume per person, which is what capacity actually looks like if response time improves and nobody leaves.

One published figure exists on this site: a B2B service provider measured 52% faster quote preparation, where the agent extracts customer requirements and prepares the information needed to create the offer. The customer name is confidential and we do not extrapolate that number to other operations, because it is a property of their data and their process as much as of the software.

Where to start

Quoting is usually the first workflow for a reason: the volume is high, the rules are written down somewhere, the outcome is defined, and the commercial payoff of response time is easy to argue internally. The Quote Agent sets out what is built before a project starts and what is configured against your catalog and pricing.

Before any of that, take ten RFQs from last week and time them honestly, from the moment the email arrived. The gap between that number and the working time is almost always the business case.

Questions

Questions we get about this.

Can it read specifications and bills of materials, not just the email?

Yes. The email body, PDF specifications including scans, spreadsheet bills of materials and word-processor documents are all read as one request, because terms in the email frequently qualify what the attachment says.

What about customers who use their own part numbers?

That is the main matching problem and it is handled by combining the description with the account's order history and your catalog. Codes that cannot be resolved confidently stop on that line with the candidates shown, and every resolution a person makes is captured so the same code resolves automatically next time.

Does it use each customer's contract pricing?

Yes, read live from the system of record at the moment the quote is prepared, not from a copied price list. Contract terms, volume breaks and active discounts therefore reflect what the account is entitled to today.

What happens with items that need engineering to price?

They stay with your team, and the agent prepares everything around them: the customer identified, the standard lines matched and priced, availability checked and the specification extracted. The engineer receives a prepared quote with one open question rather than a raw request.

Does the quote go out automatically?

Only within the limits you set. Quotes inside your policy can be sent; anything above a value threshold, below a price floor or outside the rules is routed for approval with the reason attached.

See what this would look like in your operation.

A Workflow Review traces one of your real workflows end to end and names the first step an agent could take over. No platform to evaluate first.

Start with one workflow

Fixed scope for the first workflow, agreed before we start. Expansion is your decision once it proves value.

How it works

Security, stated plainly

Where the agents run, what they can and cannot do, and exactly what we do and do not claim.

Security