How to automate construction estimating: a seven-step method
Most estimating departments do not need a new estimating system. They need fewer hours spent reading drawings, retyping quantities and re-checking addenda, and more hours spent on the bids worth winning. This guide sets out a method for automating that work one workflow at a time, without losing the traceability an estimate has to keep.
- Published
- Sources checked
- Reading time
- 9 min
- Written by
- Mirage Metrics team
Scope: estimating departments at general contractors, specialty contractors and engineering firms that price work from drawings and specifications. It covers how to choose, prepare, pilot, measure and extend the first automated workflow. It is a method, not a product review; where Mirage's own practice is described, the text says so.
Decision snapshot
- Best suited to
- Estimating managers and chief estimators whose bid volume has outgrown the team, and operations leaders asked to make estimating scale without a matching headcount.
- Inputs needed
- A map of the current workflow, a few closed projects with their drawings and quantities, the cost code structure you price in, and one recent bid already priced to pilot against.
- What can be automated
- Takeoff from drawings, subcontractor bid leveling, addendum and revision comparison, tender triage and specification review: high-volume, document-heavy work whose output can be traced to a source.
- What requires human approval
- Bid/no-bid, pricing and margin, the assumptions the drawings leave open, every flagged line, and the sign-off on the submitted bid.
- When this approach is insufficient
- When there is no stable cost code structure to land in, when the team cannot spare a reviewer during the pilot, or when the output cannot be traced back to a sheet and so cannot be checked.
The seven steps at a glance
The order matters. Each step produces what the next one needs: the map tells you which workflow to pick, the closed projects tell the pilot what right looks like, and the review rules and measures decide whether the workflow is trusted enough to scale.
- Map the estimating workflow
- Pick the first workflow
- Prepare the data: closed projects and cost codes
- Run a pilot on a bid you have already priced
- Define review and sign-off
- Measure the result
- Scale to the next workflow
Which estimating tasks can be automated?
Most of them can be prepared; few can be decided. The distinction matters more than any tool: an automated output is useful when a person can accept it quickly against its source, and useless when the person has to redo it.
| Task | Degree | What that means in practice |
|---|---|---|
| Quantity takeoff from drawings | Produced for review | Every quantity traced to its sheet and detail; flagged lines checked by an estimator |
| Addendum and revision comparison | Produced for review | The list of changed quantities and sheets; the estimator reprices what moved |
| Subcontractor bid leveling | Produced for review | Each quote mapped to scope, gaps and exclusions flagged; the award stays with the team |
| Tender triage for bid/no-bid | Prepared | A summary against your criteria; the go or no-go is a decision |
| Specification review against drawings | Prepared | Conflicts flagged with the clause and the sheet; the RFI is the team's |
| Pricing and rates | Data prepared only | Quantities land in the cost codes; rates, assemblies and judgment stay with people |
| Bid strategy, margin, qualifications | Not automated | A judgment about the client, the market and the risk |
| Handover to operations | Record kept | The traced quantities, leveled scope and addenda log travel to the project team |
Readiness checklist
Seven things to have before the first pilot. None of them is software.
- A written cost code structure, with your own additions
- Three closed projects of the type you bid most, with drawings as issued for bid and quantities as priced
- One recent bid already priced, held back for the pilot
- A named reviewer with time during the pilot
- Measurement conventions written down where the drawings leave a choice
- The import format of your estimating system, or an agreed Excel layout
- A baseline: estimator hours per bid and bids per quarter, for the last few bids
Step 1: Map the estimating workflow
Write down every task between receiving an invitation to bid and handing a won job to operations: who does it, which documents it reads, what it produces and how long it takes.
Do it with the estimators, on real bids, not from a process chart. The point is to find where the hours actually go, and the people doing the work know that better than any diagram.
| Task | Reads | Produces | Who |
|---|---|---|---|
| Bid/no-bid | Invitation, scope, schedule, contract terms | A go or no-go decision | Estimating manager |
| Takeoff | Drawings, specifications | Quantities by cost code | Estimators |
| Subcontractor bid leveling | Subcontractor quotes, scope of work | A leveled comparison | Estimators |
| Pricing | Quantities, cost database, quotes | A priced estimate | Estimators, manager |
| Revisions and addenda | Addenda, revised sheets | Updated quantities and price | Estimators |
| Review and submission | Priced estimate, qualifications | The submitted bid | Manager, leadership |
| Handover | Won bid, assumptions, quotes | Budget and scope for the project team | Estimating, operations |
Then add two columns the team fills in over a few weeks of bids: the hours each task takes per bid, and how often it is redone because something changed. Those two columns decide step 2.
How Mirage starts
Mirage's own deployments begin the same way. Its first two steps are Observe, where engineers spend time with the operational team before writing anything, and Map, where workflows, exceptions, data sources, manual checks and escalation points are mapped with that team.
Step 2: Pick the first workflow
Choose one workflow that is high in volume, heavy in documents and produces output a person can trace back to its source, and leave judgment-heavy work such as pricing and bid strategy with people.
Three criteria separate a good first workflow from a bad one.
- High volume. It happens on every bid, often several times. A task done once a quarter will not repay the setup.
- Document heavy. The hours go into reading and transcribing drawings, specifications or quotes, which is the part software reads well.
- Traceable output. Every result can be checked against the sheet, page or line it came from. If an estimator cannot check a number, they will redo it, and nothing is saved.
| Workflow | Volume | Documents | Traceable output | Verdict |
|---|---|---|---|---|
| Quantity takeoff | Every bid | Drawings, specifications | Yes, to sheet and detail | Strong first candidate |
| Revision and addendum comparison | Every addendum | Successive drawing sets | Yes, to the changed sheet | Strong candidate, often paired with takeoff |
| Subcontractor bid leveling | Every trade package | Quotes in many formats | Yes, to the quote line | Strong candidate |
| Tender triage for bid/no-bid | Every invitation | Invitation, scope, terms | Yes, to the page | Good second workflow |
| Specification review | Every bid | Specifications, drawings | Yes, to the section | Good second workflow |
| Pricing and rates | Every bid | Cost database, quotes | Partly | Prepare the data, keep the decision |
| Bid strategy and margin | Every bid | Few | No | Keep with people |
Takeoff is a common first choice because it is where the hours concentrate. At John Cockerill's Morocco operations, takeoff on 2D drawings took 3 to 4 days per project before Mirage, and up to 15 days once supplier costing was included. Piping takeoff now takes 80 to 90% less time, with engineers reviewing the output rather than measuring.
Step 3: Prepare the data: closed projects and cost codes
Gather a few closed projects with their drawings, priced quantities and final quantities, and fix the cost code structure every automated output must land in.
Closed projects
Pick projects the team knows well, of the type you bid most, with the drawing set as issued for bid, the quantities as priced and, where you have them, the quantities as built. They are the answer key: they show how your team measures, which conventions it follows where the drawings leave a choice, and where its own numbers moved after award.
Cost codes
Automated quantities are only useful if they land in the codes you price and report in. In North America that is often CSI MasterFormat, which the Construction Specifications Institute describes as the construction industry's standardized language for project documentation, organizing specifications into work results that contractors can price, procure and build. In the UK it is often an NRM structure, and NRM 2 notes that when bill of quantities items are codified, "it allows reconciliation and any necessary transfers and adjustments to be made to the cost plan". Write down the structure you actually use, your own additions included, before any software is configured.
What not to wait for
Clean historical data is not a precondition. A few well-understood projects are worth more than years of uncoded exports, and the pilot in step 4 shows which gaps actually matter.
- A few closed projects of the type you bid most, with the drawings as issued for bid
- Their quantities as priced, and as built where you have them
- Your cost code structure, with your own additions written down
- Your measurement conventions where the drawings leave a choice
- One recent bid already priced, held back for the pilot
Step 4: Run a pilot on a bid you have already priced
Run the chosen workflow on the documents of a recent bid your team has already priced, then compare the output with the team's figures line by line and record the cause of every difference.
A bid already priced gives you a reference for every line, which a live bid does not. The comparison shows where the automated output agrees with the team, where it is wrong, and where it is right and the team was not.
Illustrative example, invented figures: a pilot compared with a bid already priced
| Line | Team | Pilot | Variance | Cause | Status |
|---|---|---|---|---|---|
| CIP concrete, footings | 412 CY | 409 CY | -0.7% | Rounding in the team's takeoff | Explained |
| CMU, 8 in | 6,890 SF | 7,120 SF | +3.3% | Pilot measured elevator shaft walls the team carried as an allowance | Review |
| Hollow metal doors | 142 EA | 138 EA | -2.8% | Pilot missed four doors on enlarged plan A-401; the line was flagged low confidence | Pilot error |
| Waterproofing, below grade | 3,240 SF | 3,240 SF | 0% | None | Match |
| Reinforcing, #5 bars | 18,240 LF | 18,240 LF | 0% | None | Match |
Each difference gets a cause: a pilot error to fix, a team convention the system has to learn, or a team error the pilot caught. Record the last kind as carefully as the first.
Keep the pilot narrow: one workflow, one bid, the full set of documents that bid actually used. Mirage's proof of concept is run on the client's own documents, one week after the first meeting, and a bid already priced is a good choice for it.
Step 5: Define review and sign-off
Decide who reviews which output, what they check and who signs off, before any automated output is used on a live bid.
| Output | The reviewer checks | Signs off |
|---|---|---|
| Quantities | Every flagged line, and a sample of unflagged lines against their sheets | Lead estimator |
| Leveled subcontractor bids | Exclusions, missing scope and mismatches flagged on each quote | Estimator on the package |
| Addendum changes | The list of changed quantities, against the revised sheets | Lead estimator |
| Bid/no-bid summary | That the summary matches the invitation on scope, dates and terms | Estimating manager |
| Submitted bid | Pricing, qualifications and margin | Estimating manager or leadership |
Two rules keep the review honest. Nothing reaches a bid without a source a reviewer can open, and nothing is used until a named person has accepted it. A reviewer who cannot trace a number will measure it again, and the hours come straight back.
Start with a larger sample of unflagged lines than seems necessary, and reduce it as the variance measured in step 6 shows where the output can be trusted.
Step 6: Measure the result
Record a baseline before the pilot, then track estimator hours per bid, bids per quarter and quantity variance after review on every bid that uses the automated workflow.
| Measure | How to record it | What it tells you |
|---|---|---|
| Estimator hours per bid | A log per bid, split by task: takeoff, leveling, addenda | Whether the workflow saves time, and on which task |
| Bids per quarter | Bids submitted, and bids declined at bid/no-bid | Whether the saved hours turn into capacity |
| Quantity variance after review | Lines changed by the reviewer, and the size of each change | Where the output can be trusted and where it needs work |
| Flags that mattered | Flagged lines that turned out to be real issues | Whether the flags point reviewers at the right lines |
| Time from addendum to updated price | Hours between receiving an addendum and repricing | Whether late changes still make it into the bid |
Win rate is tempting and misleading on its own: it moves with the market, the client and the price, not only with estimating. Read it over several quarters, next to bids per quarter, rather than as the pilot's verdict.
Step 7: Scale to the next workflow
Once the first workflow is trusted on live bids, extend it to more project types and trades, then move the same team to the next workflow on the list from step 2.
Trusted means the variance after review is stable and the reviewers have stopped re-measuring unflagged lines. Extend first to adjacent work, more project types and more trades, then choose the next workflow with the same three criteria. Addendum comparison and subcontractor bid leveling are natural next steps after takeoff, because they reuse the same quantities and the same sources.
Keep the record together as you go. The traced quantities, the leveled subcontractor scope and the addenda log of a won bid are what the project team needs at handover.
How Mirage does it
Mirage deploys in five steps, in this order: Observe, Map, Build, Run, Expand. Observe: engineers spend time with the operational team before writing anything. Map: workflows, exceptions, data sources, manual checks and escalation points are mapped with the team. Build: the system is built around the constraints observed on site. Run: it runs in production with human supervision and feedback loops from day one. Expand: once a workflow is trusted, the same team moves to the next one.
The first workflow goes into production in 4 to 6 weeks; workflows spanning several systems can take longer and are scoped at the first meeting.
Questions readers ask
Where should an estimating department start with automation?
With one workflow that is high in volume, heavy in documents and produces output that can be traced to its source. Quantity takeoff and addendum comparison usually qualify; pricing and bid strategy do not, and stay with people.
Do we need clean historical data before automating estimating?
No. A few closed projects the team knows well, a written cost code structure and one bid already priced are enough to start. The pilot shows which gaps in the data actually matter.
How do you test estimating automation without risking a live bid?
Run it on the documents of a bid your team has already priced, compare the output with the team's figures line by line, and record the cause of every difference before anything touches a live bid.
What should we measure in an estimating automation pilot?
Estimator hours per bid by task, bids per quarter, the share and size of the lines the reviewer changed, and how many flags were real issues. Record the same measures on a few bids before the pilot as a baseline.
Will automating estimating replace estimators?
It replaces the reading, transcribing and re-measuring, not the estimator. Pricing, assumptions, bid/no-bid and sign-off stay with people, and every automated output needs a named reviewer.
Which cost code structure should automated takeoff use?
The one you already price and report in, with your own additions written down: often CSI MasterFormat in North America, an NRM structure in the UK. Output that lands in a different structure creates a mapping task instead of removing one.
How long does it take to automate the first estimating workflow?
It depends on the workflow and the systems involved. In Mirage's deployments, a proof of concept runs on the client's own documents one week after the first meeting, and the first workflow goes into production in 4 to 6 weeks; workflows spanning several systems take longer and are scoped at the first meeting.
Sources and scope
Checked on . Scope: estimating departments that price work from drawings and specifications. The method and the pilot example are Mirage's own; the example uses invented figures. John Cockerill's figures come from its case study on this site.
- Construction Specifications Institute, standards overview: MasterFormat described as the construction industry's standardized language for project documentation, organizing specifications into work results contractors can price, procure and build
- RICS, NRM: New rules of measurement: the three NRM volumes, NRM 1 to NRM 3, for cost planning, detailed measurement and maintenance works
- RICS, NRM 2: Detailed measurement for building works, 2nd edition UK (PDF): section 2.2, on codified bill of quantities items allowing reconciliation with the cost plan
Related on this site
- AI for estimating managers: the same method from the estimating manager's side, with a pilot you can run
- AI for quantity surveyors: measurement, variations, valuations and final accounts
- AI quantity takeoff for contractors: the usual first workflow
- Subcontractor bid leveling: a natural second workflow
- Plan revision comparison: what an addendum changed before you re-measure
- Build, buy or forward deploy: the three ways to get an automated workflow
- Hire an estimator or automate takeoff?: the decision before this method, with a cost-per-bid example
- Estimating under pressure: what to do this week, before any automation
- Automation ROI calculator: the pilot's figures, turned into payback
Plan your first estimating workflow
Mirage Metrics deploys AI agents for construction estimating teams, one workflow at a time. Tell us which task costs your team the most hours and we will send back a plan for automating it.
Prefer to talk it through? Book a call