FOR ESTIMATING MANAGERS
AI for estimating teams: your estimators price, the agents measure
AI for estimating means agents that take the takeoff, the subcontractor bid leveling and the addendum checks off your estimators, with every quantity traced to its sheet, detail and specification reference. Your team keeps the bid/no-bid call, the pricing and the sign-off. In production at OCBM and John Cockerill, and piloted on a bid you have already priced.
Running at OCBM and John Cockerill.
THE REALITY
Bid volume is set by the market and the size of the estimating team by the budget, so the gap is closed with evenings or with bids left unpriced. Takeoff, leveling and addendum re-checks take the hours that should go to choosing which bids to chase and how to price them.
THE ESTIMATING WEEK
Where the estimating team's hours go, task by task
Most of an estimating department's week is reading documents and moving quantities between them. That part can be prepared by agents and checked by people; the decisions that win or lose the bid stay with the team.
Task
What Mirage prepares
What stays with your team
Bid volume and bid/no-bid
What Mirage prepares
Tender triage: each invitation read on arrival, with the scope, key dates, documents received and the requirements that matter to your criteria summarized in one place.
What stays with your team
The decision to chase a bid, and how much of the team it gets.
Takeoff
What Mirage prepares
Quantities from PDF, DWG and IFC sets, scanned or vector, each traced to its sheet, detail and specification reference, with confidence flags, ready for the import format of your estimating system.
What stays with your team
Checking the flagged lines, the assumptions the drawings leave open, and pricing.
Subcontractor bid leveling
What Mirage prepares
Each subcontractor quote mapped to your scope, with exclusions, missing scope and quantity mismatches flagged on the line they affect.
What stays with your team
Follow-up with the subcontractors, and the award.
Revisions and addenda
What Mirage prepares
Each addendum compared with the previous set: what changed, in which quantities, on which sheets.
What stays with your team
What the change does to the price, and whether to qualify the bid.
Specification review
What Mirage prepares
Specification requirements read alongside the drawings, with conflicts between the two flagged before pricing.
What stays with your team
The clarifications and RFIs you decide to send.
Handover to operations
What Mirage prepares
The traced quantities, the leveled subcontractor scope and the addenda log kept together, so the project team starts from the same record the bid was priced from.
What stays with your team
The handover meeting and the budget the project team takes on.
WHICH TASK FIRST
Which estimating task should you automate first?
Six questions, in order. The first task that answers yes to all of them is the pilot; most estimating teams land on takeoff, with addendum comparison and bid leveling as the second and third because they reuse the same quantities.
Does it happen on every bid?
Takeoff, addenda and bid leveling do. Bid/no-bid and handover happen once per bid but take far fewer hours, and pricing is a judgment, not a reading task.
Do the hours go into reading and transcribing?
Yes for measuring sheets, re-reading an addendum in full and keying quotes into a leveling sheet. No for pricing, strategy and client calls, which stay with people whatever you automate.
Can every output be checked against a source?
A quantity against its sheet and detail, a leveled line against the quote, a changed quantity against the revised sheet. An output a reviewer cannot trace will be redone, and nothing is saved.
Is the same work repeated bid after bid?
The same assemblies, trades and drawing conventions let the measurement rules be captured once. Work that changes completely each time keeps the automated share low.
Can someone review during the pilot?
The pilot compares the agents' output with a bid your team already priced, line by line. Without a named reviewer the comparison never happens and the result cannot be trusted.
Can the output land where you price?
Excel, the import format of your estimating system or your cost database. If the answer is a new tool your estimators must open, the hours come back in a different place.
When the strain shows as bids declined for lack of time, start with takeoff. When it shows as addenda re-read in full, start with revision comparison. When it shows as evenings spent on quotes in twelve layouts, start with bid leveling.
WHAT MIRAGE DOES
What changes for the estimating manager?
With Mirage, AI agents take off the set, level subcontractor bids and compare each addendum with the previous set, and the estimating team reviews the flagged lines and prices the work.
Takeoff traced to the sheet
PDF, DWG and IFC, scanned and vector sheets in one pass. Every quantity carries its sheet, detail and specification reference, with a confidence flag where the set is unclear, so a reviewer checks the flagged lines instead of measuring the set again.
Bids leveled, addenda compared
Subcontractor quotes are mapped to your scope with exclusions and gaps flagged on the line they affect. Each addendum is compared with the previous set, and the estimator gets the list of quantities that changed rather than a new set to read.
Into the system you already price in
Output goes to Excel, to the import format of your estimating system, or straight into your cost database. Mirage connects to Procore, Autodesk Construction Cloud, Sage, SAP, Oracle and Microsoft Dynamics, and feeds your own material catalog and labor units rather than replacing them.
THE OUTPUT
What does an addendum check look like?
An illustrative extract with invented figures: the bid-set quantity, the addendum quantity, the sheet each one came from, and a flag wherever an estimator needs to look.
| Item | Bid set | Addendum 2 | Unit | Source | Flag |
|---|---|---|---|---|---|
| Slab on grade, 6 in, 4,000 psi | 64,200 | 64,200 | SF | S-101 | No change |
| Reinforcing, #5 bars, foundation walls↳ Addendum 2 thickens the north foundation wall on S-201. The 1,650 LF difference is listed for the estimator to reprice; nothing else on the sheet moved. | 21,400 | 23,050 | LF | S-201, Addendum 2 | Changed |
| Hollow metal doors, 3 ft x 7 ft | 142 | 146 | EA | A-601, Addendum 2 | Changed |
| Roof insulation, polyiso↳ Addendum 2 changed the roof insulation thickness in the specification, while detail 3/A-501 still shows the old one. Flagged as a conflict before pricing. | 38,600 | 38,600 | SF | A-501, specification | Spec conflict |
| Structural steel, W12x40 | 2,115 | 2,115 | LF | S-301 | No change |
THE PILOT
How do you measure an estimating pilot?
Run it on a bid your team has already priced, so every line has a reference, and compare the two line by line. Record a baseline before the pilot starts, then track the same four measures during it.
Estimator hours per bid
Logged per bid for takeoff, leveling and addenda, for a few bids before the pilot and for every bid during it.
Quantity variance after review
How many lines the reviewer changed, and by how much. It tells you where the agents need work and where they can be trusted.
Bids per quarter
Bids submitted and bids declined at bid/no-bid, against the same quarter before. Capacity shows up here first.
Flags that mattered
Of the lines flagged for review, how many were real issues. Too few flags hide risk; too many waste the reviewer's time.
With Mirage, the proof of concept runs on your own documents one week after the first meeting, and the first workflow goes into production in 4 to 6 weeks. The pilot is fixed price, and you can stop after it.
BEFORE YOU CHANGE ANYTHING
Our process already works. Why change it?
Often you should not change it, only take load off it. If your team prices every bid it wants to chase, on time and without evenings, and the quantities hold up at award, the process is doing its job and the case for automation is weak. The case gets stronger when the strain shows somewhere else: bids declined for lack of time, addenda read again in full, the same quantities measured twice, or one estimator holding a method nobody else can run.
How to test it without losing your controls
Start from a bid you already priced
Mirage runs the same documents and you compare its quantities with your team's line by line. Every difference has a reference: your own estimate.
Keep your controls exactly as they are
The reviewer, the pricing and the sign-off do not move. Nothing reaches a bid until a named person on your team has accepted it.
Start narrow
One trade or one document type first, the one that takes the most hours. A narrow scope gives a clear answer.
Write the pass mark down first
Set thresholds for the four measures above before the pilot starts, and decide in advance which result means you keep the current process.
When to keep the current process
- A few bids a month, priced on time without evenings: the setup will not pay back.
- Clean vector sets, moderate volume and a team happy to run a takeoff tool themselves: a self-serve product is the quicker start.
- Work that changes completely from one bid to the next, with no repeated measurement rules to capture.
- Nobody can spare the time to review output during the pilot: the comparison is what makes the result worth trusting.
Will it disrupt the bids you are pricing now?
It should not, and the way to make sure is to keep the current process running until the new one has proved itself. Nothing is migrated and nothing is replaced: Mirage writes into the tools your team already prices in.
Phase
When
What happens to your bids
Status
Proof of concept
One week after the first meeting
Mirage runs on a sample of your documents. Your live bids are not touched.
Published Mirage step
Pilot
Fixed price, you can stop after it
Runs in parallel on live or recent bids. Your estimators keep pricing the usual way; the output is compared, not used.
Published Mirage step
Switch
Trade by trade, once its pass mark is met
Output lands in Excel or the import format of your estimating system, so the pricing workflow around it stays the same.
Method agreed with you
Run
From the first workflow in production, 4 to 6 weeks in
A person reviews flagged lines from day one, and corrections your team makes become rules the agents apply from then on.
Published Mirage step
Training
There is no new estimating software to learn. Reviewers learn to read the flags and the sheet references on each line, during the pilot itself, on bids they know.
Going back
The current process is never switched off during the pilot and nothing is migrated, so stopping means carrying on as before. You can stop after the pilot.
What it does cost
Time, not disruption: a few hours with your estimators to capture your measurement rules, and a reviewer during the pilot. Plan the pilot outside your busiest bid week.
PROOF
Running in estimating teams today
OCBM prices its bids from quantities Mirage's agents measure, John Cockerill's Morocco operations run Mirage's piping takeoff in production, and Rouandi runs an estimating workflow on live projects. In all three, people review the output before it is used.
OCBM
3 agents in production
Measured quantities feed straight into bid pricing
Every quantity traceable to its source drawing
John Cockerill
80 to 90% less time on piping takeoff
Takeoff that took 3 to 4 days per project is now reviewed, not measured
Fabrication bill of materials ten times faster
Rouandi
One estimating workflow in production on live projects
Every output reviewed by the team before use
Each correction fed back into the system
Send us a bid you already priced
Choose a recent bid your team has taken off and priced. We run the same documents, you compare our quantities with yours line by line, and the differences show where the hours would go. We answer within two business hours.
How a deployment runs: Observe, Map, Build, Run, Expand. See the method
Prefer to talk it through first? Book a call
FAQ
Frequently asked questions
MORE FOR ESTIMATING TEAMS