ServicesPricingBlogAbout Book a demo
← Back to Kratu AI Labs
AI Estimating Engine  ·  India (NBC 2016)

A fire BOQ takes weeks by hand.
Watch it take minutes.

Upload a fire-protection drawing and the engine returns a priced, NBC 2016 Bill of Quantities, on your own rate card, exported tender-ready as Excel + PDF. Your engineer still reviews and signs off; the weeks of grind simply disappear.

Real output. Open it yourself.

Not a mockup. These are the actual artifacts the engine produced from our reference test drawing: the live dashboard result, and the tender-ready PDF and Excel, every line and every review note included. Open the full files and inspect them, then book a demo to see the same on your own drawing.

The review notes, in both files

Every assumption the engine made on this drawing, in its own words, so your engineer knows exactly what to confirm before the tender goes out. Here: the industrial hazard class assumed, a drawn pump that Table 7 does not call for, and device counts raised to the building quantity.

Six stages. Weeks of grind, gone.

This is the exact path your estimator walks by hand over 3–6 weeks, the engine runs it end-to-end in minutes. You keep the judgement; it kills the grind. Every stage stays checkable, and the internal rules and thresholds stay under the hood.

Kratu Estimate, core workflow
// The engine's core workflow, the stages below explain what each part does.
1

Ingest & normalize

CAD drawings. DWG or DXF, are converted into clean vector geometry the engine can reason over. A drawing that arrives as a PDF is rendered and read visually instead, so a PDF is no longer a dead end. No manual redrawing either way.

Manual: hours → Engine: seconds
2

Detect components

Every sprinkler, detector, call point, hydrant, hose reel and extinguisher drawn as a symbol is counted. Pipework, valves and fittings are then sized to code from those counts and the building.

Manual: days → Engine: seconds
3

Measure & take off

Counts taken off the drawing, pipe runs and quantities sized from them, in metric. Areas are read from what the drawing states, never guessed from pixels: the takeoff that devours the most estimator-hours of any job.

Manual: 1 to 2 weeks → Engine: seconds
4

Apply NBC 2016

NBC 2016 applied to the takeoff: the occupancy group, its Table 7 row, sprinkler and pump demand, risers, alarm and detection. The same way on every job; the full hydraulic calculation stays with your engineer.

Manual: days → Engine: seconds
5

Assemble & cost

Items roll up into a structured BOQ, priced against your own rate card, tender-ready Excel + PDF, your brand on it.

Manual: days → Engine: seconds
6

Review & sign off

Your engineer reviews, adjusts any rate, and approves, the professional accountability stays exactly where it belongs: with a human.

The judgement stays yours

No CAD drawing? A PDF works too, either kind.

A PDF reaches the engine two ways, and it handles both. When your client hands you an architectural values brief as a PDF, the engine reads it and extracts every value exactly as stated. When what you have is the drawing itself as a PDF: exported from CAD, or scanned from a print, the page is rendered and read visually, so you are not stuck waiting for a DWG that may never arrive.

The honest part: a PDF carries less than the CAD file it came from, and a poor scan carries less again. Where the sheet will not settle a dimension, the engine returns a preliminary scope flagged for review rather than inventing a number, so you can see exactly what it could and could not read. Working from an Excel values sheet instead? We'll tailor the engine to your format on request.

Priced to NBC 2016 Table 7

The building is typed to its NBC 2016 occupancy group (A residential and hotels, B educational, C institutional, D assembly, E business, F mercantile, G industrial, H storage, J hazardous) from the drawing's own words, and its own Table 7 row, by height and area, decides what is priced: sprinklers, hose reel, wet riser or down comer, yard hydrant, manual alarm, detection, public address, underground and terrace tanks, and the pump set. Floors are counted above ground, basements are sized level by level, from the basement areas the drawing states where it gives them. A building whose use is not stated is priced as mixed use and flagged, never quietly priced as residential.

Every assumption is shown, not buried

Each BOQ carries its review notes: a hazard class assumed, how the built-up area was read, a drawn pump that Table 7 does not call for. It reports what was read off the drawing and what was derived, with a confidence score. And a single floor sheet cannot under-price a whole building: where the devices drawn are fewer than the floors and area stated need, the BOQ is priced at the building quantity and says so, line by line.

Scope-honest by design: it flags low-confidence rates ("manual review recommended") and refuses to invent whole-building items from a partial drawing, so what your engineer signs off is trustworthy, not a black-box guess. That's the difference between a demo toy and a system you can bid real tenders on.

The payback math, on your numbers

Manual estimating in India isn't just slow, it's revenue left on the table. Here's the honest picture, and what flips the day this engine is yours.

The pain today
✕
3–6 weeks per BOQ

Drawing review → hydraulic calc → takeoff → rate analysis → drafting. Every single tender crawls.

✕
Tenders you turn away

A senior estimator clears only a handful a month, the rest you simply can't turn around in time to bid.

✕
One estimator, one speed

Your best estimator is a ceiling you cannot buy past, they can't go faster, they take leave, and if they resign the method walks out with them.

What changes with the engine
✓
Minutes, not weeks

A priced, NBC 2016 BOQ in minutes, about a day end-to-end once your engineer has reviewed it.

✓
Bid on everything

Quote every enquiry that lands. Each extra BOQ you can now turn around is ~₹1.5L+ of billing you couldn't capture before.

✓
Own it, runs forever

Your server, your rate card, your brand on the output. No per-estimate fee, no seat licences, no vendor lock-in.

The payback
At a conservative ₹1.5L per BOQ, just +2 extra BOQs a month pays off the ₹25,00,000 engine in
~8 months
… then it works for free, forever.

That ₹1.5L is deliberately conservative, standard commercial BOQs bill ₹1.2–2L, high-occupancy (hospital / mall / IT park) ₹2–3.6L, specialised (data centre / plant) ₹3.6–5.6L. Plug your own numbers in:

If it wins you…Extra billing / yearPayback on ₹25L
+1 extra BOQ / month (ultra-conservative)₹18,00,000~16 months
+2 extra BOQs / month (conservative)₹36,00,000~8 months
+6 extra BOQs / month (turning away work today)₹1,08,00,000under 3 months

Nobody is displaced. A good fire estimator is the hardest hire in this trade, and the engine does not replace one, it produces the first-pass priced BOQ that your engineer then reviews and signs. What changes is how many tenders that same judgement can reach: takeoff, rate analysis and drafting stop being the ceiling, so the tender list you could never turn around in time becomes one you can.

And unlike a hire, the method is written down. Your rate card, your assumptions and your output format live in the engine, so they stay in the business when people move on.

Three ways to run it

Pricing below is specific to the AI Estimating Engine only, other systems are scoped separately.

Start on the monthly. It is the smallest commitment that still puts the engine on your own drawings, it is an operating expense rather than a capital sanction, and it needs no server of yours to run. Owning it outright is what you graduate into once the throughput is proven on your own work, not where you begin.

EngagementBest forInvestment
Launch PilotProve it first, full setup & configuration on your real requirements. One-time offer per business.from ₹3,00,000 one-time
Managed System
most firms start here
We host, run & keep improving it, an operating expense instead of a capital sanction, and no infrastructure of your own to maintain.₹3,00,000 onboarding
(waived if upgrading from Pilot within 15 days)
+ ₹99,999 / month
Build & Own ItOwn it outright on your server, far cheaper over its life, and payable across build milestones rather than as a single cheque.₹3,00,000 onboarding
(waived if upgrading from Pilot within 15 days)
+ ₹25,00,000 one-time
special introductory offer, ₹40,00,000
≈ 22 months of the monthly, then yours

Put that outright figure in proportion before it frightens anybody. Counting the one-time onboarding, a managed client's cumulative spend passes ₹25,00,000 at about 22 months. So owning it is roughly two years of what you would be paying anyway, except that at the end of those two years the engine is yours, on your own server, with nothing further to pay. A firm that estimates for many years never stops paying on the monthly; that, and not the sticker, is the argument for owning it.

And it is not a single cheque. Build & Own It is payable across build milestones: scoping, configuration against your own rate card, and handover, so the spend tracks delivered work rather than landing up front. The Launch Pilot starts at ₹3,00,000 and is a one-time offer per business; upgrade to Managed System or Build & Own It within 15 days and the ₹3,00,000 onboarding fee is waived.

Honestly, so you can trust it
  • It produces a first-pass priced BOQ; a qualified engineer still reviews and signs off. It removes the grind, not the professional accountability.
  • Accuracy tracks input quality, a clean CAD drawing or a complete brief yields the best output, and it tells you when it's unsure rather than guessing.
  • Quantities are sized to code from the devices counted and the building; it does not replace the hydraulic calculation or the final layout, which stay with your engineer.
  • Built to NBC 2016 / INR / metric. (An NFPA / USD / imperial build for other markets is separate.)

See it run on your drawing.

The fastest way to believe "weeks to minutes" is to watch your own project become a priced BOQ in front of you, we'll hand you the Excel + PDF on the call.