Contractor-side QS · Consultancy · Estimating
The typing, not the judgement.
Taking off is measurable work and re-keying is not work at all. BuiltUp does the measuring and the transcription, grades what it measured line by line, and hands you a document you can check, in the client's own workbook where there is one. What it will not do is form a view. That is still the job.
No card. Three takeoffs and three AI estimates on the free plan, enough to run it against a bill you have already measured.
How the work reaches you
Two readers, and they want opposite things.
If you are contractor-side, the constraint is the return date and the number of tenders you can get out before it. Time saved is real money and you can have it as such.
If you are in consultancy, time saved is not the pitch, and pretending otherwise insults you. Where the fee is a function of hours, a tool that halves a task halves the invoice. What you can use is capacity, so the same headcount prices more tenders, and defensibility, so every figure carries its working and every automated step is recorded. This page is written for both, and it says which is which rather than blurring them.
Either way the input is the same two things: a drawing set to measure, and somebody else's document to fill in.
Take off from the set, with the confidence attached
Quantities come back with HIGH, MEDIUM or LOW against each and a note recording what the figure was read from. It is instructed never to average, never to pad, and never to invent a room it has not seen. Where a drawing carries scaled geometry it is read directly; where it does not, the estimate is capped at MEDIUM and the reference it used is named. Units are constrained to m², ft², linear metres, item, hour and day, nothing else can come back.
Price the client's schedule in the client's schedule
Where the work arrives as a spreadsheet with an empty rate column, it is filled in place. Their sections, their item references, their wording and their order, at their true row numbers. Nothing is re-sorted. Their formulas are left alone and recalculate on open, and their original is stored byte-identical and never written to.
Hold three figures apart
The comparison view puts BuiltUp's figure next to your own and next to the final account. The two variances are never averaged and there is a test in the product asserting that no combined figure is ever produced, because one of them measures the method and the other measures how the job went. A percentage against a zero baseline returns nothing rather than 0% or infinity. Lines present in the final account but absent from the specification are surfaced as probable variations rather than absorbed into the total.
Bind it to a term contract where there is one
Import your own NSR, M3NHF or bespoke schedule, bind a job to a term contract carrying the agreed uplift or tendered discount, and coded lines price at the scheduled rate with the published base rate shown next to the adjusted one. An SOR rate is treated as all-in and never runs through the material, labour and markup engine, because the schedule already deems it to include everything necessary to do the job.
Nothing here is a black box. Every line carries the material, the labour and the extras it was built from, inside its own section, so it can be read down the way a bill is read down, and disagreed with a line at a time.
Where the model is used, and where it is not.
If you are RICS-regulated you may need to record what a tool did on your behalf. This is that list. It is drawn from the code, not from a marketing position, and the second half is the one that matters.
- The model reads the drawings, the photographs and the schedule
- It proposes quantities, scope wording and line items
- It proposes a schedule code where a schedule is bound
- It drafts a method statement and a risk assessment
- And then the things it is never trusted with:
- It is never trusted with the item reference or section of a schedule row
- Row identity is parsed by rule and tested, because a reference that drifts silently re-points the client's number at the wrong line
- It is never trusted with the price of a catalogue material. That comes from your catalogue
- A schedule code it returns is looked up in your imported library and dropped if it does not verify
- A code whose unit is not dimensionally compatible with the measured line is rejected outright
- The risk score on a RAMS is recomputed server-side from the likelihood and severity actually chosen
- The arithmetic in your client's workbook is never touched. Their formulas are skipped and reported as preserved
- A material only matches when the names genuinely overlap. A partial match keeps its own label and price rather than adopting a catalogue entry's
Nothing BuiltUp produces is competent advice and nothing it produces is signed. A RAMS is a draft until a named person with the competence to approve it does so, and the approval records who and when. You remain the accountable person on every output.
The four a surveyor actually uses.
Chosen rather than listed. The full capability list is on the features page.
- Variances kept apart
- Was the method right? Did the job go to plan? Those are two questions and two numbers, and the product refuses to average them. If you have ever been handed a single "accuracy" figure that turned out to be those two things blended, you already know why this is here.
- Schedules of rates, under your own licenceUltimate · Beta
- BuiltUp supplies no rate data at all. You bring the schedule, and the import is blocked until you record a licence and attest that it permits loading into third-party software. That is not a legal hedge. It is how this market works: contractors cannot buy the M3NHF schedule directly, they receive it from the landlord. Reissued editions import alongside the old one and live contracts stay on the edition they were let under.
- An audit trail that survives an adjudicationPro
- Correspondence, changes, contracts, quotes, invoices, files, site attendance and the programme are copied and fingerprinted into a dated pack, hashed over a canonicalised snapshot, with a verify endpoint that recomputes the hash to prove the record is intact. It also reports what is missing from the record, while there is still time to fix it.
- The document goes back as the document
- A priced return in the issuing surveyor's own workbook, at their own row numbers, is quicker for them to check than anything you could rebuild. That is worth saying plainly: it makes your tender the easy one to assess.
One real job
A client sent a 33-item schedule of works. It came back priced to within about 6% of their own estimator.
Not a drawing. Their spreadsheet, with a rate column left empty. BuiltUp priced all 33 items and returned the same document: their sections, their item numbers, their wording, in their order. Their estimator had already priced it by hand. The two totals were within about six per cent of each other.
That is one job, not a benchmark, and we are not going to present it as one. What lands on your schedule depends on your own rates, and you should check it against your own figures before it reaches a client, which is exactly why it comes back in a document you can read line by line.
How the handback worksWhat it will not do for you.
- It is not a measurement standard.
- It does not claim NRM2 or CESMM4 compliance and it does not produce a bill in a standard's prescribed format. It measures, it labels in the wording an SOR item would use, and it grades itself. If the contract says SMM7, the mapping is still yours.
- Automatic coding is weaker on very large schedules.
- Where a bound library is 1,200 items or fewer the whole schedule is put in front of the model. Above that it retrieves a slice using phrases from the brief. The code records a void-relet test where that retrieval found none of the 24 codes required. Verified codes are still verified. The risk is codes that never get proposed at all, so check coverage on a large published schedule before you rely on it.
- No connector to Procore, Buildertrend or CoConstruct.
- Clients, projects and scope items import from a CSV using a template BuiltUp gives you, with duplicates flagged before anything is written. Supplier price lists import from a spreadsheet or a PDF. Everything else is an export and a column mapping.
Questions this comes with.
- I bill by the hour. Why would I use something that makes me quicker?
- You would not, if that were the only effect. The argument for a consultancy is capacity and defensibility: more tenders priced with the headcount you already have, and a record showing what was measured, what confidence that measurement carried, and which steps were automated. If your firm is contractor-side, the time is the point and you can take it at face value.
- How do I record what the tool did, for our AI policy?
- The list above is the boundary, taken from the code. Beyond that, every action in the product is logged, RAMS carry an approver's name and role, and evidence packs are hashed and verifiable. What we cannot yet publish is the model provider, whether customer data trains anything, the hosting region and the retention period. That needs engineering to confirm, and we are not going to write a reassuring paragraph in the meantime. Ask, and you will get whatever is actually established.
- How accurate is it?
- Honestly, it depends on your rates and on how good the document is. One real job came back within about 6% of the client's own estimator. One job, described above, not a benchmark, and we will not restate it as an accuracy rate. What is promised instead is that it shows its working: every measurement graded, everything unpriceable left blank with a reason, and every schedule code verified before it prices. Test it on a job you already know the answer to.
Run it against a bill you have already measured.
Your figure exists. Put the two next to each other and see where and why they differ.
No card. One active project, three takeoffs and three AI estimates.
Or set it up with us, 30 minutes, screen-shared