Tender season: what an RFQ actually costs your estimator

Ask a commercial builder what their constraint is and most will say estimator capacity. Ask what an estimator actually spends the day doing and a large share of it is reading, not estimating.

This is a worked example of where those hours go, and what changes when the reading is handled before it reaches a person.

The setup

A Brisbane commercial builder. Around $40M turnover, three estimators, work split between fitout, light commercial new build, and some government-adjacent projects through panel arrangements.

Tender volume runs at roughly 18 invitations a month. They bid about 6. They win about 1.5.

That ratio is normal and it is also the whole problem. Two thirds of the tender documents that arrive are for jobs they will not bid, and the effort of working out which two thirds is spent before anyone knows.

What an RFQ actually costs

A tender invitation arrives as an email with links or attachments. A typical commercial package is somewhere between 40 and 300 pages across multiple documents: the invitation itself, drawings, a specification, a bill of quantities or schedule of rates, preliminaries, special conditions, and a prequalification questionnaire. Addenda arrive later and change things.

Before an estimator can decide whether to bid, someone has to establish:

• What the actual scope is, and which trades are included or excluded
• The closing date, and whether it has moved
• Site constraints, staging requirements, and access limitations
• Whether the contract conditions are acceptable, or contain something the business will not wear
• Insurance, prequalification, and compliance requirements
• Whether there are ambiguities that need an RFI before pricing is possible

For a mid-sized package that is two to four hours of reading before a bid or no-bid decision can be made responsibly.

At 18 invitations a month, that is roughly 45 to 55 hours of estimator time spent on triage. Around two thirds of it, call it 35 hours, is spent on tenders they will not bid.

Thirty-five hours a month is not far off a full week of one estimator, spent reading documents to reach the answer "no".

The bit that actually hurts

The hours are the visible cost. The expensive part is the timing.

Because triage is slow, the no-bid decision often arrives late. A tender that lands on the 3rd might not get properly read until the 9th. If the answer is no, five days of the bid window are gone and the estimator could have spent them on a tender they would have won.

Worse, the reverse happens too. A good opportunity sits in the queue while someone reads a package that was never viable, and by the time it surfaces there is not enough time to price it properly. The business either rushes it, or lets it go.

The result is a pipeline where the decision about what to bid is driven substantially by what happened to get read first, rather than by which jobs the business is best placed to win.

What they tried

Reading the invitation email and skimming. Fast, and it works until the day the exclusion that kills the job is on page 140 of the specification rather than in the summary. Everyone who has done this has been burned at least once.

A junior doing first-pass triage. Reasonable, and many builders do it. The limit is that the judgement required to spot a problematic special condition is exactly the judgement a junior does not have yet. It moves the hours rather than removing them, and it introduces a risk of missing the thing that matters.

A go / no-go checklist. Genuinely helps, and worth having regardless. But the checklist still has to be filled in by someone who has read the documents, so it improves the quality of the decision without reducing the cost of reaching it.

Bidding less. The instinct when estimators are drowning. It usually reduces the wrong end: businesses bid the jobs that are easy to assess rather than the ones with the best margin.

What document processing actually does

The useful framing is not "AI writes the tender". It is that a machine reads 200 pages in under a minute and produces the structured brief a human would have spent three hours producing.

Concretely, for each incoming package it extracts:

• Scope items, itemised, with the trades each falls under
• Closing date and time, plus any addenda that have changed it
• Exclusions and inclusions stated in the documents
• Contract type and any non-standard special conditions, flagged for review
• Insurance, prequalification, and compliance requirements
• Ambiguities and contradictions, which is often the highest-value output, because a contradiction between the drawings and the specification is an RFI that has to go out early or it delays everything

That lands in front of the estimator as a structured summary with page references, so the judgement calls still get made by a person, but they get made in fifteen minutes instead of three hours, and on day one instead of day six.

This is what the tender and RFQ processing work looks like in practice, and it usually connects to whatever the business already runs, whether that is Procore, Aconex, Buildertrend, or a shared drive and Outlook.

What changes

Triage stops being the bottleneck. The 35 hours a month spent reading toward "no" collapses to a few hours of reviewing summaries.

The no-bid decision moves to day one. This is the real win. Knowing on the day it arrives that a tender is not viable means the whole bid window goes to something else.

RFIs go out earlier. Ambiguities surfaced on day one become questions asked while there is still time for an answer, rather than assumptions priced in under deadline.

Bid selection improves. When assessing a tender is cheap, the business can assess all of them properly and choose based on fit and margin rather than on reading order. For most builders this matters more than the hours saved.

Addenda stop slipping through. Automated re-reads catch what changed between revision A and revision C, which is a common and expensive miss.

What it does not fix

It does not price the job. Quantities, rates, subcontractor pricing, programme, and risk allowance are estimating work and stay with the estimator. Anyone selling you automated pricing on commercial work is overselling.

It does not make the go / no-go decision. It produces the brief the decision is made from. The judgement is still yours, and it should be.

It is not perfect at reading drawings. Text-heavy documents, specifications, and schedules extract reliably. Marked-up drawings and hand-annotated PDFs are harder, and anything scanned at low quality will need a person.

It does not replace relationships. A large share of commercial work comes through knowing the client and being invited to the right jobs. Reading tenders faster does not change who invites you.

How to test it on your own tenders

You do not need to take this on faith, and you should not.

Take five tender packages you have already processed, including at least two you decided not to bid. Have them run through document extraction, then compare the output against what your estimator concluded at the time.

Three questions tell you whether it works for your document mix:

1. Did it find the same scope items your estimator found?
2. Did it flag the thing that drove your actual decision? If you no-bid because of a special condition, did it surface that condition?
3. Did it invent anything? This is the one to watch. An extraction that is confidently wrong is worse than no extraction, so check the page references.

If it passes on your own historical tenders, it will hold on new ones. If it does not, you have lost an afternoon rather than committed to anything.

If you want to run that test, send us five packages and we will process them and walk through the output with your estimator. Real documents, your judgement on whether it earned its place.