AI ESTIMATING · M&E
Short answer: Only check a takeoff that shows its working. Every line should carry the drawing sheet and revision it came from, and anything unclear should be flagged with a reason. Clear the flagged lines first, spot-check counted lines against the sheets, check the high-value items in full, and have your estimator sign off before a single rate goes in.
Why can't I just trust the AI's quantities?
For the same reason you wouldn't price a junior estimator's takeoff without looking at it. The numbers might be right, but your margin depends on them, and one missed submain or a floor of FCUs counted twice comes straight out of your profit.
The aim of checking isn't to redo the takeoff. It's to confirm the draft quickly and catch the lines that need your judgement. That's only possible if the draft is built to be checked.
What should an AI-drafted BQ line look like?
Each line needs enough information that your estimator can open the right sheet and confirm it in seconds. At minimum:
| Field | Example | Why it matters |
|---|---|---|
| BQ ref | E1/03 | Matches the consultant's item numbering |
| Description | Lighting point wired in 1.5mm² PVC cable in PVC conduit | Follows the consultant's wording |
| Quantity and unit | 46 pts | In the BQ's own units |
| Sheet and revision | E-11 Rev B | Tells you exactly where to look |
| Status | Counted, or Check | Separates confident lines from ones needing you |
| Reason (if Check) | Outdoor unit position differs between M-03 and roof plan A-09 | Tells you what to decide |
If a draft gives you a total with no sheet reference, there's nothing to check against. Your estimator will end up recounting it, and the AI has saved no time.
How do I check an AI takeoff quickly?
Work in this order. It puts your attention where the risk is.
- Confirm the drawing register. Check the list of sheets and revisions the draft used. If it counted from a superseded revision, stop and fix that first.
- Clear every flagged line. Read the reason, open the sheets it names, and decide. Correct the quantity or confirm it, and note your decision.
- Check high-value lines in full. DBs, submains, chillers or VRF outdoor units, switchgear, solar inverters. These are few in number and large in value, so check each one.
- Spot-check the counted lines. Pick a sample from each sheet, for example one floor's lighting points, and count them yourself. If your count matches, move on. If it doesn't, check that whole sheet.
- Check for missing items. Scan the BQ for item headings with no quantity, and check the specification for items that aren't drawn.
- Sign off. Your estimator approves the takeoff. Only then are rates applied.
What kinds of lines get flagged, and how should I treat them?
| Flag reason | Typical example | What to do |
|---|---|---|
| Sheet not to scale | Submain length on a riser diagram marked "not to scale" | Measure from a scaled plan, or allow from site route and note it |
| Drawings disagree | FCU count differs between layout and schedule | Decide which governs, or raise a tender query |
| Symbol unclear | Fitting type not in the legend | Check the specification or ask the consultant |
| Equipment position differs | Condensing unit on roof plan vs ACMV layout | Confirm position before pricing pipe runs |
| Poor scan | Faint symbols on a reduced PDF | Request a clean copy or count manually |
A flagged line isn't a failure. It's the AI doing what a careful junior should do: putting a question to the person who decides, instead of guessing.
How do I check measured items like cable and trunking?
Measured runs carry more risk than counted points, because scale, routes and allowances all affect the result.
- Check the scale on the sheet. Confirm the draft used the stated scale and that the PDF wasn't reduced.
- Check one run end to end. Pick a long cable tray or trunking run and measure it yourself.
- Check vertical runs. Rises between floors and drops to equipment are often shown only on schematics or not at all.
- Check your allowances. Make sure your firm's usual allowances for wastage, terminations and bends were applied as your rules say, not added twice.
How do I know the AI is reliable before using it on a live tender?
Test it on tenders you've already priced. That's the only fair comparison, because you already know the right answer.
In a UkurBQ setup, this is built into the handover. The first past tender is run together with your estimator, line by line, adjusting the setup wherever it differs from how you measure. Two more past tenders are then run blind, with no changes, and compared with your own BQs. Your team keeps quoting the usual way until those blind runs match.
After that, checking is a routine rather than a leap of faith: each line shows its source, and anything that can't be confirmed is flagged instead of guessed.
FAQ
How long should checking take?
It depends on the size of the tender and how many lines are flagged. The point is that checking a traced draft is much quicker than counting from scratch, because you're confirming rather than measuring.
Who should do the checking?
The estimator who would normally price the job. They know your measuring rules and will spot anything that looks wrong for that type of building.
What if my spot-check doesn't match?
Check the whole sheet, correct the lines and note what went wrong. If it happens repeatedly on the same type of drawing, the setup needs adjusting.
Can the AI apply our rates once we've signed off?
Yes. Your own rate sheet is applied line by line in the consultant's BQ layout, inside your own account, after sign-off.
Try it on a tender you've already priced
Send one past tender. We return a draft BQ with every quantity traced to its sheet, so you can compare it with your own. Free, under NDA.
Book a free BQ test