Ask anyone who’s actually put a tender submission together where the time goes, and they won’t say pricing. They’ll say re-entering the same project history for the third or fourth time that week: once into the estimate, once into the submission template, once into whatever format the client’s procurement portal demands.
That’s the actual bottleneck. Not the technical writing. Not even the pricing itself, most quantity surveyors can price a job reasonably fast once the numbers are in front of them. The slow part is getting the numbers in front of them in a usable form.
That distinction matters, because it changes what actually needs fixing.
Where the Time Actually Goes
Break down an average tender and the hours split roughly three ways: pulling historical rates from old jobs to price the new one accurately, chasing subcontractor quotes that come back in five different formats, and reconciling scope changes that land after someone’s already started pricing. None of that is technical skill. It’s data retrieval and reformatting, done under deadline pressure, usually by one or two people instead of the four or five roles that theoretically exist for it on paper.
There’s also a version control problem sitting underneath all of this that rarely gets named directly. Three people working from their own copy of the same rate schedule. One gets updated with a revised subcontractor quote, the other two don’t, and nobody notices until the client’s procurement team flags that two line items in the same submission don’t reconcile. That’s not a training problem or a competence problem. It’s what happens when the same information lives in five different spreadsheets instead of one place.
If the actual pricing work takes a day, the total submission often takes three. Almost none of that extra time is spent thinking. It’s spent looking things up and typing them into a different place.
The person absorbing most of that cost is usually the quantity surveyor, or whoever’s drawn the short straw of pulling the submission together this time round. If you want the fuller explanation of what we mean by an AI agent versus a standard AI tool, and how that distinction applies more broadly, we’ve covered that separately in what AI agents actually mean for WA engineering teams. Worth reading first if this is new territory.
When Tenders Overlap, the Real Cost Shows Up
Most of the time this is manageable. One tender, one deadline, someone gets it done even if the last two days are rougher than they should be. The real cost shows up when two or three tenders land in the same window, which happens more often than project pipelines suggest it should.
At that point the same one or two people are doing historical rate lookups for three different jobs at once, and the corners that get cut aren’t the technical ones. They’re the double-checking. A rate that’s twelve months out of date makes it into a submission because there wasn’t time to verify it against the most recent job. Nobody catches it until the numbers come back wrong on a won contract, which is a considerably more expensive place to find the error than during pricing.
What ChatGPT and Copilot Actually Do Here, and What They Don’t
ChatGPT for tender submissions and Copilot’s estimating construction workflows inside Microsoft 365 are useful for exactly one part of this: writing. Turning a set of bullet points into a coherent technical narrative, tightening up a methodology statement, catching inconsistent terminology across a fifty-page document. Genuinely useful, worth using.
A methodology statement that’s been rewritten four times by four different people usually reads like it. Running it through ChatGPT or Copilot once at the end, for tone and consistency rather than content, catches the seams other proofreading tends to miss.
What they don’t do is pull last year’s rate for civil earthworks out of a project folder from eighteen months ago, or notice that a subcontractor’s quote came in without GST when every other line item includes it. That’s not a drafting problem. It’s a retrieval and reconciliation problem, and it’s the part that actually eats the other two days.
Where an Agent Changes the Actual Mechanics
An agent built for this pulls historical rates automatically instead of someone searching through old job folders, keeps one live version of the cost sheet instead of five people’s separate copies, and flags a scope change that needs repricing instead of relying on someone catching it before submission. None of that replaces the quantity surveyor’s judgement on the final number. It means they’re making that judgement with the right numbers already in front of them, rather than spending two of the three days finding them.
We’ve built this specifically around ai estimating software construction for engineering and construction estimating, not as a generic pricing tool with a chatbot bolted on.
We don’t have a published number for tendering specifically yet. Elsewhere in our work, the same underlying approach cut new project setup time by 70% for McGill Engineering, which is the same category of problem: pulling the right historical information into a new job automatically instead of manually rebuilding it each time.
Mining Tenders Carry Their Own Weight
Mining tender automation in WA carries its own wrinkle: compliance documentation and safety case requirements that don’t show up on a standard civil tender. Contractors submitting to major WA producers know exactly how much heavier that paperwork gets. The retrieval problem is the same, just with a longer list of historical documents to pull from, covered in more detail in what AI automation looks like for WA mining services contractors.
Retrieval and reconciliation, not pricing skill, is what’s actually costing the two extra days. Fix that part and the technical writing and pricing logic, the parts a quantity surveyor is actually trained for, take about as long as they should.
If tender turnaround is what’s slowing your team down, it’s worth seeing what the retrieval and reconciliation side of this actually looks like once it’s automated properly.