Estimating in construction has a bottleneck that never moves: an enquiry arrives, and somebody experienced has to turn it into numbers.
The enquiry itself is rarely tidy. It might be a paragraph in an email, a schedule attached as a spreadsheet, or a photographed page from a specification. Whatever the format, the same job follows. Read what has been asked for. Work out which product in the catalogue it actually means, when the wording never matches the catalogue wording. Find the right price for that customer, because the discount depends on who is asking. Add the labour, because the material on its own is not a quote. Then lay it out, total it, and send it.
That is an hour of a senior estimator's day, per enquiry, on work that is mostly matching and arithmetic. It is also the work most likely to be done late, because it competes with live jobs, and the enquiries that wait are the ones that are lost.
We build the AI inside APSIS Business Components, a construction software platform. Quoting is one of the capabilities we built into it, and it takes the enquiry all the way from arrival to a priced document.
An enquiry enters as an email body or an uploaded file. The AI reads it and pulls out the lines being asked for, each with its description, quantity, unit, and any brand or specification hint the sender gave. From there each line is matched against the account's own catalogue and its own labour rates, scored and ranked, so the estimator sees the best matches rather than a search box. Customer-specific supplier discounts apply automatically, labour is priced from productivity rates rather than guessed, and the result is a quote that can be revised, snapshotted and issued as a PDF.
The enquiry is read, whatever shape it is in. An email body, a spreadsheet, a photographed page. What comes out is a set of lines, each with the description as written, the quantity, the unit, and any brand or specification the sender mentioned in passing.
Nobody writes descriptions the way the catalogue does. A request for "100mm blocks, 7N" has to become a specific catalogue line with a real code and a real price. The system proposes candidates from the account's own catalogue, each scored and ranked, and the estimator confirms rather than searches. That is the difference between assisting an estimator and replacing their judgement, and it is deliberate.
Labour is priced, not guessed. Labour rates carry productivity per day and a task factor, so the labour on a line is calculated from the same evidence as the material rather than added as a percentage at the end.
The price depends on who is asking. Customer-specific supplier discounts apply automatically, so the number that comes out is the number for that customer, not a list price somebody then has to adjust.
Every issue is a revision. Quotes are lettered, with the totals and the lines snapshotted at the moment of issue and the document stored. When a client says "your quote said", there is a record of exactly what it said.
A quote answers one enquiry. A bill of quantities prices a whole job, and it is where estimating gets genuinely difficult, because the structure has to hold up under scrutiny from a quantity surveyor.
The estimating side of the platform models that properly rather than approximating it in a spreadsheet. A bill holds items, and an item holds subitems, so a single measured line can carry its own build-up. Each subitem sits against a trade and a category, and prices separately for labour, material, plant and subcontractor, with subcontractor uplift handled as its own rate rather than folded into a margin. Quantities can be overridden where the measure and the reality differ, and the override is visible rather than silently replacing the number underneath.
Above all of that sit the project measures a developer actually asks about: gross external area, gross internal area, net internal area, and the number of units. So the same bill answers both questions, what the work costs and what it costs per square metre, without anyone rebuilding it in a second spreadsheet.
Bills can be issued as a platform standard, so a business prices in a consistent structure across every job instead of each estimator inventing their own layout.
Quoting does not sit on its own. It is one of the capabilities we have built into the APSIS platform over more than a year, alongside procurement document processing, drawing and schedule extraction, specification comparison, contract risk analysis, bid writing and document search.
They share the same spine, which is why each one arrives faster than the last: the same accounts and permissions, the same document handling, the same audit trail. A drawing extracted in one module is available to a question asked in another.
Read how we build the AI inside the platform →, or see what the same approach does with procurement paperwork.
Estimating capacity is the ceiling on how much work a contractor or supplier can chase. Every enquiry that waits three days is an enquiry someone else has already answered, and the ones that get declined for lack of time never appear in any report.
Taking the matching and the arithmetic off an estimator does not replace them. It means the same estimator covers more enquiries, and the judgement they are paid for goes on the lines that need it rather than on looking up block sizes.
Tell us where your team loses the most time. We will tell you honestly whether AI pays there, what it takes to build, and what we have already delivered for businesses like yours.
We use cookies to measure how the site is used and, if you allow it, for advertising. Nothing is set until you choose. Privacy policy.