How AI tender reading actually works
Feeding a 200-page tender to a model isn't the hard part. Knowing which clause to flag is.
The bid rarely gets lost on price. It gets lost on a line item nobody double-checked.
Contractors rarely lose a tender because their number was too high. They lose because one line inside that number was wrong: a quantity copied straight off the consultant's BOQ, a rate carried over from last year's job, a unit read as "per item" when it should have read "per set." The total looked competitive. The bid still bled money, or got flagged out during evaluation over a technicality nobody caught in time.
These aren't rare slip-ups. The same five mistakes show up on tender after tender, because the same pressure shows up every time: a two-day submission window, a stretched estimating team, and a BOQ running hundreds of lines that everyone assumes someone else already checked.
Most contractors price against the quantities the consultant issued in the tender documents, which is normal enough under a two-week window. The mistake is treating those figures as fact instead of someone else's estimate. Consultant-issued BOQs carry their own errors: a slab counted twice across two drawing revisions, a quantity that never got updated after a late design change, and rounding that quietly shifts a concrete or earthworks figure more than it should.
If the quantity is wrong and you price it as given, you either underprice the real work and eat the difference on site, or overprice it and lose the tender to whoever caught the error. Take off the 15 to 20 highest-value lines independently, since they usually carry most of the contract sum, compare them against the tender figures, and raise a technical query on anything meaningfully out while the clarification window is still open.
The second habit is reusing last job's rate build-up because rebuilding one from scratch under tender pressure eats time nobody has. Steel and cement prices move, and so do labor costs, with Iqama renewal fees and accommodation costs shifting from one contract cycle to the next. A rate that held up a year ago can already be off by the time you're pricing a tender today.
Treat this as a standing discipline: pull at least two current supplier quotes for your major material lines before you finalize a bid, and treat any rate older than 60 to 90 days as provisional until it's refreshed. Build labor rates from current basic wage plus GOSI, accommodation, and transport, not a flat daily figure carried forward from an old project.
Preliminaries, site overhead, mobilization, temporary facilities, project management staff, and insurance get underpriced because they aren't exciting measured items. They get a lump sum scribbled in during the final hour, based on gut feel rather than actual project duration and site conditions. Provisional sums get the opposite treatment: contractors leave them exactly as issued, assuming they're the client's problem, without checking what's included.
This mistake is purely mechanical and still costs money every tender season: pricing against the wrong unit. Tiling priced per m2 next to skirting priced per linear meter, cladding measured per m2 of elevation instead of actual panel area, mechanical items marked "per set" that got priced "per item," which can run two or three times off. It happens fast when an estimator is pricing hundreds of lines in two days and pattern-matches off a similar item from the last job.
The fix is procedural: whoever finalizes pricing reads the unit column against the description for every line, not just the ones that look unusual, and a second estimator spot-checks unit consistency across similar trades before the file goes out. Ten minutes of cross-checking beats an hour of rework after the contract is signed.
Every BOQ has a handful of items worded loosely on purpose: "make good all disturbed areas," "as directed by the engineer," "allow for minor works as instructed." Priced literally, these lines cost almost nothing. In practice they're where scope creeps back in after award, because the wording gives the consultant room to instruct more than you assumed. Pricing them at zero because the description is vague isn't pricing risk, it's ignoring it.
Flag every ambiguously worded item during the tender clarification period, even if the window is short. A written answer from the consultant becomes your reference if a dispute comes up later. Where clarification isn't possible before submission, build a measured allowance into those specific lines instead of a blanket markup across the whole BOQ, and keep a note of your reasoning. That note is what lets you defend the number later instead of caving during negotiation.
None of this needs new software or a bigger team. It needs someone reading the BOQ a second time after the first pass feels done: the units, the rates, the preliminaries, the vague lines nobody wanted to think about the night before submission. That second read is usually the only thing between a bid that wins and a bid that wins and then costs you money for the next year.
Want this handled for Riyadh 12211 or the rest of the Kingdom? Talk to Tender Mind AI.
Learn more →