AI takeoff software that reads the whole tender before it measures

Why should AI takeoff software read the whole tender?

The AI takeoff software worth using is the one that reads the whole tender before it measures a thing. In Estimator+ that's AI Takeoff V5, and it's the main reason we'd tell any commercial builder to have a look at us.

Think about where a quantity really comes from. The plan shows a door tag. The schedule says what the door is. A general note says every door gets a stop. And Addendum 02 deleted two of those doors on Tuesday.

Measure the plan on its own and you get a tidy number for the wrong scope. Anyone who's had a subbie ring up about a clause on page 40 of the spec knows how that goes.

What does it read before it measures?

Everything in the tender that changes the quantities. Before AI Takeoff V5 measures a single line, it reads:

• the specification
• the schedules
• the general notes
• the contract terms and the project scope
• every addendum

From there it recognises how the job's being delivered, Design & Construct (D&C) included, catches the trade exclusions and factors in the spec notes before it starts on the drawings.

Here's the sort of thing it'll come back with. It tells you a big chunk of the project is D&C, then asks whether to add those items and tag them as unmeasured, or keep them as RFIs. That one's your call, so it puts it to you.

Notes that drive quantities

Some of the biggest numbers in a tender aren't drawn at all. They're written. "Allow 14 per floor." "Door stops to all doors." A takeoff that goes looking for symbols walks straight past both.

AI Takeoff V5 counts what a note calls for. And where the documents say a note governs over a schedule, it follows the note.

Here's an invented example. The general notes on an office job say "Allow 14 access panels per floor, locations to be confirmed." The note covers Levels 1 to 6, and no panels are drawn. So the takeoff counts 14 on each of the six floors, 14 × 6 = 84 access panels, on a line of their own. Measure the plans alone and that line never exists.

Typical floors, scanned sheets and the current set

A few things that usually chew up an arvo:

• Typical floors. It measures a typical floor once and carries it to every level it covers, with the exceptions each level prints.
• Scanned and rotated sheets. It reads scans with no text layer as well as clean PDFs, so the detail the engineer scanned in sideways still gets taken off.
• The current revision. It works from the current revision of every sheet and applies each addendum before measuring, so a deleted wall or a changed door type never reaches the quantities.

That last one earns its keep in the final week before close. Tender addenda and revisions covers what happens when an addendum lands.

What's in, what's out, and what's the client's?

A takeoff that measures everything drawn will happily measure the existing slab you're not touching. So it leaves out work by others, existing work to remain and anything that isn't in the contract.

The client's own money gets its own spot too. PC sums, provisional sums and priced options are set up as their own trades, worded the way the tender words them. They carry PS and CO tags through the estimate, Review and Quotes, so nobody mistakes a provisional sum for something you measured.

Every assumption out in the open

No tender answers every question. So when the documents leave something open, the AI assumes, says so and keeps going, rather than stopping halfway through the run. You get every assumed height, extent and unit, a flag wherever two drawings disagree, and the open questions asked together at the end.

The RFIs come ready to send. Each question is addressed to the consultant who should answer it, so a ceiling height goes to the architect and a beam size goes to the structural engineer. It also lists any documents missing from the set, which is much nicer to find out now than on close day.

Why it never types a number

This is the bit we care about most. The AI never types a quantity. Every quantity is a shape on a published sheet (an outline, a line or a dot) or a pin, and Estimator+ works out the number from that shape and the sheet's printed scale.

The scale gets proven on the sheet itself, from two points and a printed length, or the printed ratio where there is one, and your own calibration is never replaced. So you're checking a shape you can see. Can ChatGPT do a takeoff? gets into why that matters more than which model's doing the reading.

Lines get filed under the standard trade list, one line per thing for the whole job, named with the drawing's code and product. Something like "CT1 - 13mm plasterboard ceiling; ceiling area". Your subbies see the product they're pricing, and nobody has to go hunting for what CT1 means.

How do you check it?

In Review mode, every measured shape carries its line's name right on the plan, so you check the takeoff off the drawing instead of flipping between a list and the sheets. An outline that's wandered off its room shows up straight away.

Need to add something by hand? The manual tools are there too, for area, length, count and wall height.

That check stays yours, and so do the price, the picks and the strategy.

The honest catch

It reads drawings as PDFs. It doesn't take off CAD or BIM models, so if you price from models, keep your model-based tool for that work.

The takeoff runs in the Claude desktop app or in Codex in the ChatGPT desktop app, with your tender folder open. ChatGPT in a browser can help with the estimate, quotes and pricing, but not the takeoff. Connecting Claude or ChatGPT walks through the four-step setup.

The takeoff comes in every Estimator+ seat, A$450 per seat per month ex GST, billed monthly with no contract. The Claude or ChatGPT plan is your own, so there's no AI bill from us. And you'll still have questions to answer at the end of each run, because some calls are yours to make.

The fairest way to judge it is on a job you've already priced, so bring one along when you book a demo and compare the AI's takeoff with yours, line by line.