Why Change Orders Eat Margin at Mid-Market Contractors
Most $40M general contractors lose 3 to 5 percent margin on change orders. The fix is workflow: clear thresholds, a 48-hour rule, and field discipline.
TL;DR. Most mid-market general contractors lose 3 to 5 percent margin on change orders, and it is not because the work is hard. It is because there is no workflow. Field crews start before the change is approved. PCOs sit in the project manager's inbox for two weeks. Subs invoice for work that was never papered. The fix is operational, not technological: tiered approval thresholds, a 48-hour documentation rule, and cost-code discipline per change. AI helps with drafting and pattern surfacing. AI does not fix the field-to-office gap.
You lose margin on change orders because three things happen on every job at a $40M general contractor. Field crews start work before the change is approved. The PCO sits in the PM's inbox for two weeks. The owner disputes the cost three months later. None of those three is a software problem. All three are workflow. This post walks through the workflow that fixes them, where AI helps, and where it does not.
Why Change Orders Are 7 to 15 Percent of Your Project Cost
Industry research from Engineering News-Record and the Associated General Contractors of America consistently puts change orders at 7 to 15 percent of total project cost on most commercial jobs. On a $5M job, that is $350K to $750K of work outside the original contract. Lose 3 to 5 percent margin on that work (industry typical) and you give back $10K to $37K per job. Run 25 jobs a year and the number is real.
The math gets worse with consequential costs. ELECTRI International research on electrical trade work shows true change order overhead can reach 19 percent against the typical 10 percent markup. That gap is crew reassignment time, stacked-trades productivity loss, PM hours chasing approval, and accounting time coding correctly. Standard 10 to 15 percent markup caps do not cover any of it. Even when the change gets approved and paid, you often net less margin than the base contract.
The field-to-office handoff is where most change order margin leaks. The conversation at the point of change is worth more than any software upgrade.
Where Margin Leaks at $30 to $50M Contractors
Five specific failure modes show up on every job.
1. Field Crews Start Before the PCO Is Approved
The owner's rep walks the site, points at a wall, says "move that two feet east." The super agrees because the work is small and the owner is on site. By Thursday the wall is moved. By Friday the PM finds out. No PCO, no T&M tracking, no cost code that maps to the change. The hours go into base contract cost or into a generic "field changes" bucket nobody can defend in an audit.
2. PCOs Sit in the PM's Email for Two Weeks
The PM has 11 active jobs and 8 PCOs in flight. Some are half-drafted, waiting for a unit price or a field quantity. By the time the PCO goes out, work is half done, the crew has moved on, and reconstructing actual hours is a forensic exercise. Construction Executive reports up to 30 percent of change orders trace back to inadequate planning or incomplete designs, meaning PMs absorb PCO load they should never have absorbed.
3. Subcontractor Change Orders Get Lost in the Email Thread
The drywall sub gets a verbal "yes, go ahead" from the super for a $3K patching add. The sub does the work. Six weeks later they invoice. Nobody on the GC side has a record. The sub gets paid because they have the email chain. The GC eats it because there is no backup PCO from the owner. This is the coordination failure described in Why Sub-Contractor Coordination Breaks at 15 Active Jobs, and it compounds across every change.
4. The Owner Disputes the Cost Three Months Later
You submit the PCO. The owner-rep verbally approves. Work proceeds. The formal change order never gets countersigned. Three months later at the owner's audit, they question the $47K item and ask for documentation. You have a verbal yes and field notes. They have a contract clause requiring written authorization. You are negotiating from weaker ground, and the work has long since been billed.
5. The Estimator Never Sees the Trend
By year-end you have closed 200 change orders across 25 jobs. The patterns are obvious in hindsight: renovations under-bid demo by 7 to 9 percent every time, concrete pours come back with at least one weather-related change. The estimator never sees it because the data lives in 25 separate job folders. Next year's bids underprice the same risks. The same loop drives Your Quoting Process Is Costing You Jobs: the field knows things the bid does not, and there is no mechanism to feed it back.
The Workflow That Actually Holds
The workflow itself is simple. The discipline of running it is the hard part. Five pieces.
1. Tiered Approval Thresholds
Define approval authority by dollar value. Most $30 to $50M GCs run something like:
| Change Value | Field Authority | Office Authority |
|---|---|---|
| Under $2,500 | Super can direct the field | PM countersigns PCO within 24 hours |
| $2,500 to $25,000 | Super flags, holds work | PM signs, owner-rep verbal okay, written within 5 days |
| $25,000 and up | Super flags, holds work | PM signs, owner countersigns before work proceeds |
The exact numbers depend on your contracts. The principle: every dollar value of change has a single person authorized to bind the company, and they are identified before the job starts. No more "I thought you said yes." Public-sector contracts already enforce this discipline. Caltrans Section 5-3 is a reasonable reference for how to write the thresholds.
2. The 48-Hour Rule
Every PCO is submitted within 48 hours of identifying the change. If it cannot be priced cleanly in 48 hours, a placeholder PCO goes in with "T&M, to be priced" and the field tracks actual hours and materials in real time under a dedicated cost code. The 48-hour rule is the highest-leverage discipline in change order management because it forces the field-to-office conversation while the details are still fresh. Wait a week and you are reconstructing.
3. Dedicated Cost Codes Per Change
Not one generic "CO" code. Every PCO over a threshold (say $1,000) gets its own cost code the moment the PCO is logged. Labor and materials flow into that code in real time. Work-in-progress reports then show change order margin separately from base contract margin. When something is slipping, you see it in week two, not at closeout. The same discipline runs through Job Costing for Contractors: What Your P&L Won't Tell You, and change order margin is where it gets tested first.
4. Standard PCO Template
One template, every job, regardless of which PM is running it. Required fields: change description, originating party, photo at the point of change, schedule impact, cost breakdown (labor, materials, equipment, sub, OH&P), and approval signature line. PMs cannot submit a PCO without all fields completed. The template lives in the project management system, not a Word doc on a laptop. The template alone is worth a margin point because it eliminates the "PCO from Tom looks nothing like PCO from Maria" problem that drives owner disputes.
5. Weekly Change Order Review
A standing 30-minute meeting, every Monday, every active job. The PM walks the open PCO log: what is outstanding, what is pending owner signature, what has been worked but not papered. Two questions get answered. "Where are we waiting?" "What is our exposure if the owner pushes back?" This is the meeting that catches the $47K dispute six weeks before it becomes one.
The Monday PCO review is the cheapest insurance a mid-market GC can buy. Thirty minutes per job catches disputes six weeks before they surface in an owner audit.
Where AI Actually Helps (And Where It Does Not)
AI is useful in three specific places in the change order workflow. Not in the places people sell it hardest.
AI Helps
Drafting PCO narratives from field notes. The super takes a voice memo: "Owner walked the site, wants the wall moved 24 inches east, we need to rework the east elevation framing, probably 8 hours, materials are 2x4s and drywall." An AI agent drafts a PCO narrative from that, suggests a cost code, and pre-fills the template. PM reviews, adjusts, sends. Saves 15 to 20 minutes per PCO. On 200 PCOs a year that is 50 to 65 PM hours freed up for the work that needs judgment.
Surfacing patterns across past change orders. Run the last 18 months of PCOs through an agent that pulls out cost categories, root causes, and originating party. The estimator gets a report: "On your last 12 renovation jobs, demo came in 7 to 9 percent over takeoff every time. Bid the next one with a contingency for that." This is the kind of analysis estimators do not have time to do manually but that pays for itself the first time it changes a bid. The underlying issue is often an RFI that never got answered fast enough, and the pattern surfacing is what makes that connection visible to the estimator.
Flagging missing PCOs against the schedule. An agent reading both the project schedule and the open PCO log flags work that should have a PCO but does not yet. "Cost code 02-200 had 47 labor hours last week against base contract. Schedule shows base demo complete on the 14th. Possible un-papered change." The agent does not decide anything. It surfaces the discrepancy so the PM can ask the super.
AI Does Not Help
Negotiating with the owner-rep. Software does not have judgment on whether to push a borderline change or eat it for relationship reasons. That call belongs to the PM and the principal.
The 48-hour discipline. No tool fixes a workflow that nobody runs. If the PM does not check the PCO log, the agent's outputs sit unread.
The field-to-office handoff. The super has to actually tell someone the change happened. Voice memos and photos help; they do not replace the conversation. The contractors who do change orders well are not running better AI. They are running a 7am huddle where the super walks through every change identified in the last 24 hours.
What This Costs to Implement
A $40M GC usually has the workflow pieces in place already (template, cost codes, meeting cadence) but the discipline is uneven across PMs. Standardizing takes about 60 days: a half-day session to align on thresholds, a week to refactor the cost code structure in the ERP, a month of holding every PM to the new template and the 48-hour rule, and a month of catching the back-sliding. No new software required.
The AI tooling on top (voice-memo-to-PCO drafter, pattern surfacer, schedule-vs-PCO flagger) is a 4-week build. Most of the work is the integration into your project management system (Procore, Buildertrend, Sage, whatever you run), not the AI itself.
The recovery math works. Tighten the workflow to recover one percentage point of margin on change order work (conservative), and at 10 percent change-order revenue that is $40K a year per $40M of base contract. Pays back in the first quarter. Everything after compounds into operating margin.
What to Do This Week
Three diagnostics, 30 minutes each. The patterns tell you which piece to fix first.
- Pull all PCOs from your last 3 closed jobs. Measure the gap between change date and PCO submission date. Median over 7 days means you have a 48-hour-rule problem.
- Look at one job's cost code structure. Count how many change orders are coded to a single generic "CO" code. More than two means you have a cost-code discipline problem.
- Ask three PMs how they decide whether the field crew can proceed with a change. If the answer varies, you have a threshold problem.
If this sounds like your Tuesday, we would be glad to talk. Granular builds focused AI tools and operational agents for $30 to $100M general contractors: voice-memo-to-PCO drafters, schedule-vs-PCO flaggers, pattern surfacers on historical change orders. Fixed price, four-week delivery. Book 30 minutes and bring one open PCO that has been sitting on your PM's desk too long. We will tell you where the margin went.
Keep Reading
- Job Costing for Contractors: What Your P&L Won't Tell You: Change order margin is the first place job costing discipline gets tested. This post covers the full picture, including what your P&L is hiding from you on the base contract side.
- Why Sub-Contractor Coordination Breaks at 15 Active Jobs: The companion operational failure mode to lost change order paper. When sub coordination breaks down, change order disputes are the symptom that shows up first in your closeout reports.