How to Automate RFQ Processing in a UK B2B Company (Without Losing the Human Judgement)

Most UK B2B sales teams do not lose deals on price. They lose deals on the speed and quality of the quote that follows the enquiry. The pipeline number on the board looks healthy, and underneath it the inbound RFQ queue is quietly stretching by a day, then by 3, until "we are managing" turns into a competitive disadvantage that nobody named.

RFQ automation exists to take the typing out of that bottleneck. Done well, it cuts the per-RFQ handling time by 90% or more, frees the team for the work that moves the deal forward, and keeps the human firmly in control of the pricing.

This is the playbook we use with UK B2B businesses in the £4M to £160M revenue band. It assumes you already have inbound RFQ volume and a manual process that is starting to bite. The anchor proof point throughout is a UK commercial flooring business that took RFQ handling from 20 minutes to under 90 seconds.

What is RFQ automation?

RFQ automation is the use of AI and workflow tooling to handle the repetitive, cross-system work that sits between an inbound Request for Quote arriving and a structured, quotable project in your system. It reads the email and attachments, extracts the project data, matches the customer record, and either creates the project automatically or flags it for a human review. The human still owns pricing and the commercial call; the automation owns the typing.

The phrase covers everything from a simple email parser to a full agentic pipeline. The pattern that consistently delivers in UK B2B is the middle of that range: a workflow that handles mixed inbound formats, scores its own confidence, and routes to a human on the edge cases.

Where the hours go in manual RFQ

The hidden cost of manual RFQ work is rarely on anyone's dashboard. It hides inside operations and sales support, in calendar fragments, and in the work that did not happen because someone was reformatting a spreadsheet at 4pm on a Friday.

The breakdown below is what we typically see for a mid-market UK B2B sales operations function before automation. Numbers shift by sector; the proportions are consistent across construction, professional services, distribution, and specialist manufacturing.

Where the time goes	Per inbound RFQ	Why it disappears
Reading the email and attachments	4 to 6 minutes	Spreadsheets, PDFs, and photos all need scanning before any data entry can start
Cross-referencing the customer database	4 to 6 minutes	Email domain, company name, and site references rarely match cleanly across systems
Typing project details into the system	6 to 8 minutes	Free-text gets retyped into structured fields, with the operator doing the format translation in their head
Quality checking the entry	2 to 3 minutes	Catches obvious errors but rarely the subtle miscodings that surface as commercial disputes later
Total per RFQ	16 to 23 minutes	Repeated every day, by the same people you would rather have on the live deals

For a team handling 10 inbound RFQs a day, that is between 2.5 and 4 hours of manual processing every working day, before anyone has actually written a quote. Salesforce's State of Sales research finds that sales reps spend roughly 28% of their week actually selling, with the rest absorbed by admin and the connective tissue across systems. RFQ handling is one of the heaviest absorbers in B2B sales operations.

The cost that does not appear on the timesheet is bigger. Slow time-to-quote means the customer is talking to a competitor while you are still typing. The quote that should have gone out on Tuesday goes out on Friday, and by Monday the buyer has signed with whoever moved first. This is the same pattern we covered in the hidden cost of manual RevOps: the most expensive RevOps work is the work that did not happen.

A 5-step playbook for automating RFQ in a UK B2B company

The five steps below are the ones we run with every RFQ automation engagement, in order. None of them is technically heroic. The discipline is keeping the human in the loop at the right points and resisting the temptation to "just automate everything".

Step 1: Map the inbound, end to end

Before any build starts, document what actually arrives. Pull a representative sample of 30 to 50 inbound RFQs from the last quarter. Tag each by format (email body only, spreadsheet, PDF, photo, mixed), by source (existing customer, new enquiry, partner referral), and by complexity.

What you usually find is that 70 to 80% of the inbound fits 3 or 4 patterns, and the long tail is genuinely awkward. The playbook automates the patterns and routes the long tail to humans by design. Trying to automate the awkward 20% on day one is the most common reason these projects stall.

Capture the baseline numbers in the same session: average handling time, error rate, and current time-to-quote. Without these baselines, no later phase can prove value.

Step 2: Build a structured intake, not a clever inbox

The automation needs a single, structured destination for every inbound. In practice, that means a dedicated RFQ inbox (Outlook, Gmail, or shared mailbox), monitored by a workflow that fires the moment a new message lands.

The trigger is the easy bit. The structure is the work. Decide upfront what a "project" looks like in your downstream system: which fields are mandatory, which are optional, which are pricing-relevant, and which are reference-only. The structured target model is what the AI extracts into. Skipping this step produces fast junk: data that arrives quickly and then needs a human to clean up before it can be quoted.

For UK B2B teams, the destination is usually a CRM (HubSpot, Salesforce), a project workspace (Airtable, monday.com), or a quoting platform. Build the structured intake first and let the storage system follow.

Step 3: Extract with confidence scoring on every field

This is the AI step, and it is the smallest part of the build in practice. The workflow routes each attachment through the right pipeline by file type: spreadsheets, PDFs, Word documents, and photos through vision AI. The combined email and attachment context goes through an extraction model that returns structured project data with a confidence score on every field.

Confidence scoring is what makes the rest of the playbook work. A high score means the field is reliable enough to auto-create. A low score means the workflow flags the specific field that needs a human eye. The IBM Institute for Business Value found that organisations capturing real productivity gains from AI are the ones redesigning workflows around AI outputs, rather than dropping AI features into manual processes. Confidence routing is that redesign for quote intake.

Step 4: Route on confidence, not on optimism

High-confidence RFQs are auto-created in the downstream system, with a confirmation notification to the team. Lower-confidence RFQs are flagged for human review, with the specific fields that need attention surfaced clearly.

Thresholds matter. Set them too high and everything ends up in human review. Set them too low and the automation silently miscodes the awkward inbound, which is worse than the manual baseline because nobody notices until a customer query surfaces it. Most UK B2B teams settle into a confidence threshold of around 0.85 to 0.90 on the critical fields (customer, site, scope) and lower on the reference-only fields.

Every routing decision should be logged. The log tells you, 3 months in, whether to tighten or loosen the threshold based on actual behaviour rather than guesswork.

Step 5: Keep the human on the pricing call

The last step is the one that gets skipped most often and matters most. The automation creates a structured, quotable project. It does not write the quote.

Pricing, discounting, payment terms, and any judgement about commercial risk stays with a human owner. The automation hands them a clean project, surfaces relevant CRM history, and gets out of the way. The team's job changes from typing customer data into a screen to making the commercial call faster, on better information.

This is the Human-in-the-Loop pattern we apply to every quote-to-cash automation. It is what makes RFQ automation safe to ship without a 6-month compliance review.

Manual RFQ vs Automated RFQ vs Hybrid (HITL)

The choice in front of most UK B2B sales operations leaders is not "automate everything" or "stay manual". It is which version of automation. The table below sets out how the three patterns compare on the things that actually matter to the business.

In practice, the hybrid pattern is the one we deploy almost every time. The fully automated version looks attractive on a slide and underperforms in operation, because the silent failures cost more than the speed gain. The manual version becomes untenable above roughly 5 inbound RFQs a day.

What's a competitive time-to-quote benchmark for UK B2B?

Time to quote is the single most useful operational metric for an RFQ workflow. It is measurable, it tracks customer experience directly, and it shows up in win rate within a quarter of any meaningful improvement.

From our work with UK B2B clients across construction, professional services, and specialist distribution, the benchmarks line up consistently:

• Over 1 week from inbound to sent quote: competitive disadvantage. The buyer is talking to a competitor while your team types. The pipeline number looks fine on the board because the lost opportunities never appear in the CRM in the first place.
• 5 working days: competitive. You are matching the market average for mid-market UK B2B. The deal stays in play, but you have no speed advantage.
• 2 to 4 working days: strong. You are ahead of the local market, the customer notices, and your win rate on contested deals begins to climb.
• Under 2 working days: leading. This is where most well-implemented hybrid automations land once they are live, and it is a defensible competitive advantage in any UK B2B sector where inbound RFQs drive the pipeline.

The supporting macro data tells the same story from a different angle. PwC's 2026 Global CEO Survey found that 56% of CEOs have seen neither revenue gains nor cost savings from their AI investments to date, and the pattern across that 56% is almost always the same: AI was deployed inside manual workflows rather than reshaping them. Time to quote did not move because the workflow did not move. For sector context, the UK Office for National Statistics tracks the labour productivity gap between UK firms and the leading OECD economies, and the admin load on UK B2B sales operations is one of the underlying drivers.

Where Human-in-the-Loop matters

There are 4 decision points in an RFQ workflow where the human stays in the loop by default. Anything else is a judgement call about your specific risk profile.

1. Confidence-flagged extraction. Any field below the agreed threshold gets surfaced for a human reviewer before the project is created. The reviewer sees the original inbound, the extracted value, and the confidence score, and confirms or corrects in one click.
2. First-time customers. A new customer record always benefits from a human eye, both for data quality and for commercial qualification. Auto-create the project, but flag the customer record for review.
3. High-value or non-standard scope. Any RFQ above a defined value threshold, or one that does not match a standard scope template, goes to a named quoter. The automation pre-populates the project; the pricing call is human.
4. Anything irreversible. Sending a quote, accepting an order, or triggering a downstream invoicing event is never auto-actioned without a human approval.

The full reasoning sits in our piece on Human-in-the-Loop design. For RFQ automation specifically, the principle is simple: take the typing away, never the judgement.

The team you already have becomes more valuable on the day the typing goes away. The same person who used to spend 4 hours retyping spreadsheets is now reviewing 10 minutes of flagged exceptions and writing better quotes in the time they got back. See the hours-saved range your own team could plausibly reclaim using our capacity calculator.

Practical takeaways

If you take 5 things away from this playbook, take these.

1. Time-to-quote is the metric that matters most. Capture it before you build anything. Over a working week is a competitive disadvantage you can measure.
2. Document the inbound before designing the extraction. 70 to 80% of inbound fits 3 to 4 patterns; the long tail is meant to go to humans.
3. Build the structured destination first. The AI extracts into a schema you define. No schema, no automation worth shipping.
4. Confidence scoring is the whole game. Every field, every time, with routing rules tuned on logged behaviour over the first 90 days.
5. The human owns the price; the automation owns the typing. Hybrid beats full automation every time we have tried, because the silent failures of fully automated extraction cost more than the speed they win.

Frequently asked questions

What is RFQ automation?

RFQ automation is the use of AI and workflow tooling to handle the repetitive, cross-system work that sits between an inbound Request for Quote arriving and a structured, quotable project in your system. It reads the email and attachments, extracts the project data, matches the customer record, and either creates the project automatically or flags it for a human review when confidence is low. The human still owns pricing, discounting, and any commercial judgement; the automation owns the typing.

Can AI handle complex quotes?

AI handles the data extraction and routing reliably across messy formats, including emails, PDFs, spreadsheets, and photo attachments. It does not handle the pricing strategy on a complex quote, and it should not be asked to. The pattern that works is to use AI for everything up to the priced line items, then hand the structured project to a quoter who applies the commercial judgement. Complexity in the inbound is fine; complexity in the pricing call belongs to a human.

What about Human-in-the-Loop (HITL)?

Human-in-the-Loop is the design principle that keeps a human owner on every decision that matters. For RFQ automation, that usually means confidence-based routing on the extraction step, manual sign-off on any quote above a defined value, and a human reviewer on edge cases the model flags. The goal is to take the typing away from the team, not the judgement. Done well, HITL gives you the speed of automation without the risk of silently miscoding the awkward inbound.

How accurate is AI for RFQs?

On well-structured inbound, modern extraction models reach the high 90s on the fields a quoter actually needs (customer, site, contact, scope, references). On messy inbound, mixed formats, and unfamiliar templates, accuracy drops, which is exactly why confidence scoring matters. The right benchmark is not 100% per field; it is the share of RFQs that flow end to end without a human correction, and the share that get correctly flagged when they need one. Both numbers should be visible from day one.

What's a competitive time-to-quote benchmark for UK B2B?

From the inbound RFQs we see across UK B2B engagements, anything over a working week of time-to-quote is a competitive disadvantage. 5 working days is competitive. Sub-2 days is leading, and is where most automated workflows land once they are live. The number that matters more than any single benchmark is the trend: a time-to-quote that is improving month over month, with a stable error rate, is the signal that the automation is doing its job.