A change order rarely turns into a fight because the form was hard to fill in. It turns into a fight because the facts, the cost, the schedule hit, and the sign-off never land in the same place at the same time.
You know the sequence. A site instruction arrives. A subcontractor says the work has changed. A vendor sends a number. Finance wants to know what it does to margin. The owner wants to know why. And you need a decision before the next pour, the next delivery, or the next crew mobilizes. Meanwhile the change register is a week behind the site, and the real exposure is sitting in an email thread nobody has reconciled.
If you run the projects, run commercial, or own the owner relationship, this is the workflow that decides whether margin moved because the job changed or because the paperwork fell behind. This guide walks a change from the field event that starts it to a priced, approved, documented order, and shows where it stalls and how to keep cost, scope, and the decision moving together.
What a good change workflow has to answer
Before you touch a tool, get clear on what a change workflow has to answer on any given day. It is really three questions. What changed, who asked for it, and which instruction or drawing backs it up. What it does to cost and to the schedule, and whether the vendor number has landed. And which changes are approved, which are disputed, and which are still exposure you have not priced.
None of those are hard in theory. They are hard because the answers live in different places. The project management tool knows the field event. The estimator's spreadsheet knows the price. The contract folder knows the entitlement position. The forecast knows none of it until someone updates it by hand. The whole job of the workflow is to put those answers together while the decision still matters, not after the forecast is already wrong.
A five-minute test on one open change
Open one change that is still live and time yourself. Can you see what changed, the latest cost number, the schedule hit, who owns the next decision, and the email or drawing behind it in under five minutes? If you are opening three systems to answer that, the workflow is hiding the exposure rather than showing it. That is the gap you are closing.
Where the time actually goes today
Most teams already have the parts. A project management tool, a cost system, document control, a folder of vendor quotes, and a change register in a spreadsheet. The time does not disappear inside any one of them. It disappears in the gaps between them.
A change starts life as a site instruction or a line in a meeting note. Someone re-types it into the register. Someone else prices it. The vendor quote lands in an inbox and sits there. The forecast gets updated on a different day by a different person, working from whatever version they happened to have. By the time the change reaches the owner meeting, half the room is arguing about which number is current instead of whether to approve it. Some exposure gets counted twice, some never gets counted at all, and nobody can say cleanly whether this month's margin move is real or just late data.
The cost of that is not only rework. It is the change that quietly expires. Notice periods pass, the entitlement weakens, and a legitimate claim becomes an absorbed cost because the paperwork moved slower than the contract required.
Walk the last twenty changes before you design anything
Do not design the ideal process on a whiteboard. Take one live project and walk the last fifteen to twenty changes from first signal to final outcome, exactly as they happened. For each one, note when the field knew, when the register knew, when it was priced, when it went to the owner, and when the forecast finally caught up.
Walking real changes shows you where they actually stall, and it is rarely the pricing. Usually it is the wait between the field knowing and the register knowing, and again between pricing the change and the owner deciding. Those two waits are where margin leaks, because during them the work often proceeds anyway. The crew does not stop because the change order is unsigned. So you are already spending the money while the entitlement to recover it is still sitting unconfirmed.
You will also find changes that never became change orders at all. A verbal instruction on site, acted on, never written up. Those are the most expensive of the lot, because the cost is real and the backup does not exist. Counting how many of your last twenty fall into that category tells you how much of the fix is process and how much is plain discipline.
A single change order, walked step by step
Here is the path a change should travel, and the point on each step where it tends to break. Read it against how your own last change actually moved.
It starts with a signal: a site instruction, an answered RFI, a field condition nobody drew, or an owner asking for something different. The break here is simple. The signal reaches a person, not a system. It lives in that person's head or inbox until they get to it, and the clock on any notice period has already started.
Then someone captures it as a change and links the drawing, photo, or email behind it. The break is that capture and backup happen at different times, when the backup happens at all. A change logged without the instruction behind it is a number you cannot defend later.
Next it gets priced, or a vendor quote is requested. This is the step teams assume is the slow one, and it usually is not. The delay is waiting on a subcontractor number, or pricing a change whose scope is still moving because the field condition has not settled.
Then the commercial review: someone checks scope and entitlement against the contract. Is this a genuine variation the owner should pay for, or is it rework the contract already covers? The break is that this judgment lives in one person's reading of the contract, written down nowhere, so no two people describe the position the same way.
Then it goes to the owner for a decision. The break is format and timing. A change submitted without clear backup, or dropped on the owner between meetings, waits for the next cycle while the work continues on site.
The owner approves, rejects, or disputes it, and the reason gets written down. The break is that disputes go into free text or into a conversation, and the next person cannot tell whether a change is dead, parked, or still being negotiated.
Finally the forecast and billing position update. The break is that this happens last, by hand, often by someone who was not in the room, so the cost report always trails the site.
| Step | Who usually moves it | Where it breaks |
|---|---|---|
| Signal arrives | Site team, PM, or owner | Lives in an inbox; the notice clock is already running |
| Capture and link backup | Project engineer or quantity surveyor | Logged without the instruction or drawing behind it |
| Price or request a quote | Estimator or quantity surveyor | Scope still moving; vendor number outstanding |
| Commercial and entitlement review | Commercial manager or quantity surveyor | Contract position held in one head, written nowhere |
| Submit to owner | PM or commercial lead | Weak backup or wrong timing pushes it a full cycle |
| Owner decision recorded | Owner or owner's representative | Disputes vanish into free text |
| Update forecast and billing | Finance or quantity surveyor | Done last, by hand; the cost report trails the site |
A worked example, from rock seam to approved order
The following is invented to show the shape of the workflow, not a real project. The numbers and timings are illustrative, chosen only to make the steps concrete.
Picture a mid-rise residential build. During bulk excavation, the crew hits a seam of rock that the geotechnical report did not flag under that part of the footprint. The site engineer photographs it and calls the PM. That is the signal, and on this project it is nine in the morning on a Tuesday.
In a workflow that holds, the engineer opens a change on the spot, attaches the photos and the inspection note, and marks it as likely exposure, not yet priced. The excavation subcontractor is asked for a number to break and remove the rock. The quantity surveyor pulls the geotechnical report and the contract to check whether unforeseen ground conditions sit with the owner or the contractor, and writes that position down in plain language on the change itself.
By Thursday the subcontractor quote lands: a figure for additional breaking, haulage, and two days of program impact. The quantity surveyor prices it, adds the contract markups, and moves the change to priced, not approved. Because the two-day schedule hit is captured on the same change, the planner can already see it pushing the next slab pour, which matters more to the owner than the dollar figure.
The change goes to the owner at Friday's meeting with the photos, the geotechnical reference, the priced backup, and the entitlement note attached. The owner asks one question, gets the drawing reference in the room, and approves it. The forecast updates that afternoon, and the two-day slip is already in the program the owner saw.
| When | What happened | Cost position | Schedule | State |
|---|---|---|---|---|
| Tue AM | Rock seam found, photographed, logged | Unknown | Unknown | Likely exposure, not priced |
| Thu | Subcontractor quote in; priced with markups | Priced | Two days flagged | Priced, not approved |
| Fri | Submitted to owner with full backup; approved | Approved value | Two days in the program | Approved |
| Fri PM | Forecast and billing updated | In the cost report | Reflected in the pour date | Closed |
Now run the same event through a broken workflow. The photos sit on the engineer's phone. The rock gets broken because the crew cannot stop. The quote lands in an inbox and waits. Three weeks later, at the monthly review, someone finds an unexplained overrun in excavation, nobody has the geotechnical reference to hand, and the notice period for claiming an unforeseen condition has quietly passed. Same event, same rock. The only difference is whether cost, schedule, backup, and the decision traveled together.
The states that decide whether your forecast is honest
The one distinction worth building carefully is exposure. A change register that lumps everything into open and closed tells finance almost nothing. The states have to separate what the owner has actually agreed to pay from what you are hoping they will.
Keep approved value apart from what is priced but not yet approved, and keep both apart from likely exposure you have flagged but not yet quoted, and apart again from disputed items the owner is contesting. That separation is what lets finance forecast honestly instead of guessing, and it is what stops a forecast from looking precise and being wrong.
| State | What it means | How to treat it in the forecast |
|---|---|---|
| Likely exposure | A change is coming but no price exists yet | Carry as a flagged risk, never as agreed value |
| Priced, not approved | A number exists; the owner has not agreed it | Show as pending, held apart from approved value |
| Approved | The owner has agreed to pay | Move into the confirmed forecast and billing |
| Disputed | The owner is contesting entitlement or amount | Hold separately with the reason; do not net it away |
The moment those four are visible at a glance, the monthly review changes character. Instead of arguing about whether a number is real, the room can talk about the two or three disputed changes that actually need a decision.
The first version that helps
You do not need a new platform in month one. You need one queue that holds every change event next to the four things people argue about: what backs it up, what it costs in money and time, where it stands in those states, and who owns the next move. Everything else can wait.
Build it from what you already have. A shared table, fed by exports from the systems that hold each piece, is enough to start. The point is not the technology. The point is that a single place now answers the five-minute test, and every change has one current version instead of several competing ones.
What the first change queue should hold
At the start, keep it to the event and its source with the instruction or drawing behind it, the cost and schedule impact with the vendor quote attached or flagged as missing, and a clear state drawn from the list above. Resist the urge to add fields nobody will maintain. A queue that captures three things reliably beats one that captures fifteen and goes stale by the second week.
| Change event | What backs it | Where it stands | Next owner |
|---|---|---|---|
| Owner-requested layout change | Instruction email and revised drawing | Priced, not approved | Owner review |
| Unforeseen ground condition | Field photos and inspection note | Likely exposure, not yet priced | Commercial manager |
| Vendor substitution | Supplier quote and spec comparison | Possible credit, in finance review | Quantity surveyor |
Pricing backup that survives an owner's scrutiny
A change gets approved faster when the owner can see how the number was built, not just what it is. Pricing backup is the difference between a change the owner signs in the room and one they park for their own quantity surveyor to pick apart over the following month.
Good backup ties the price to the change. It carries the subcontractor or supplier quote, a comparison against the original scope so the owner can see exactly what is new, the labor and equipment behind any self-performed work, and the markups applied strictly as the contract allows. When a credit is due, because a change removes work as well as adds it, that shows too. Owners trust a number more when it is not always in the contractor's favor.
This is where discipline pays off later. If the owner disputes a change three months on, the file that priced it is the same file that defends it. A number you cannot trace back to a source should never reach the owner in the first place, because you will not be able to stand behind it when it is questioned. The commercial decision on what to submit, and what markup to claim, stays with your project team. This guide takes no view on your contract terms; the entitlement read is yours and your advisors'.
Shortening the owner approval cycle
The wait between a priced change and an owner decision is often the longest part of the whole path, and it is rarely about the money. It is about how the change arrives. A single change emailed to an owner between meetings competes with everything else in their inbox and waits for the next site meeting anyway.
Two things shorten the cycle. The first is packaging. Bring changes to a standing point on the owner meeting agenda, each with its backup already attached, so the owner reviews a clean list instead of hunting for context. The second is thresholds agreed up front. Small changes under an agreed value can move on a faster track, so a hundred-dollar change does not sit in the same queue as a hundred-thousand-dollar one. Both are agreements about format and authority, and both are worth settling before the project is busy rather than during a dispute.
The measure that matters here is days from priced to decision. If that number is climbing, the problem is almost never the owner being difficult. It is usually that changes are reaching them without enough to decide on, or in a form that is easy to defer.
Keep cost and schedule impact on the same change
Most change registers track money and forget time. That is a mistake, because on many projects the schedule impact of a change matters more to the owner than the cost, and the two are not separable. A change that adds two weeks to the program can cost more in delay than in direct work, and if the time impact is not captured when the change is priced, the claim for it is usually gone.
Every change should carry its schedule impact next to its cost, tied to the program the owner has seen. When the planner can see a change pushing a milestone at the moment it is priced, the owner conversation is honest and early. When the schedule hit only surfaces weeks later as a missed date, it becomes a separate argument with weaker backup. Keeping both on the same change is also what feeds a clean profitability forecast, because margin moves with the time as much as the cost.
The data and systems this touches
The data you need is spread across project management, accounting or ERP, estimating, contract folders, procurement, and field reporting. You do not have to integrate all of it to start. You have to agree one thing first: which system owns each state. Where does the cost estimate live? Who owns the moment a change becomes approved? Where does the forecast get updated, and from what?
Start with exports and the shared queue for the change states that cause the most rework, and connect deeper only once the team trusts what the queue says. Two neighboring workflows feed this one directly. An item in RFI and submittal routing often becomes a change order here, and many changes begin as a site report or field issue. Wiring those to the same source of truth means a field condition raised in the morning is already a logged change by the afternoon, instead of being re-keyed three days later from memory.
Where AI helps, and where it must not
AI earns its place on the preparation, not the decision. It can read a long field note and pull out what changed, compare a vendor quote against the original scope and flag the line items that are genuinely new, draft the owner-facing explanation of why a change is justified, and catch a submission that is missing its quote before it wastes a review cycle. Each of those saves an hour a person would otherwise spend assembling, and none of them decides anything.
What it must not do is decide entitlement, approve the change, or stand in for commercial judgment. The pattern that works is plain. AI assembles and drafts; an accountable person reviews and decides. If a number or a justification cannot be traced back to a source, it does not go to the owner. Everything submitted to an owner, and every commercial approval, stays with your project team. The machine speeds up the paperwork around the decision; it does not make the decision.
What tells you it is working
You will feel it before you measure it, but measure it anyway. Three numbers tell the story. Days from a change first appearing on site to it being priced. Days from priced to an owner decision. And the share of your forecast that is approved value rather than unpriced exposure.
When those three move in the right direction, the monthly forecast review stops being an argument about data and starts being a decision about the project. That is the signal the workflow is holding. If they stall, the numbers usually point straight at the step that is breaking, which is more useful than a general sense that changes feel slow.
Common traps
The first trap is treating the register as the workflow. A register captures what was decided; it does not show you what is still exposure, and a change only becomes visible once it is already resolved. The second is letting entitlement live in free text, so no two people read the contract position the same way and every review reopens the same argument.
The third is trying to integrate every system before anyone trusts the queue, which turns a two-week fix into a six-month systems project that stalls under its own weight. The fourth is counting disputed and likely exposure as if it were approved, which makes the forecast look precise and wrong at exactly the moment the owner is relying on it. The fifth is tracking cost and ignoring time, so the schedule claims evaporate. Keep the states honest, keep time next to money, and the rest gets easier.
How Ubisar would build this with you
In week one, we would take one project and walk its last fifteen to twenty changes from first signal to commercial outcome, then stand up a single change queue: what backs each one, its cost and schedule impact, its state, its exposure, and who owns the next decision. In weeks two and three, we would connect the least data that keeps that queue current, drawn from project management, accounting, estimating, and field reporting, and add AI where it drafts and flags under a person's review. By week four, your forecast review should run off that one queue.
We work month to month, one workflow at a time, as AI, data, and tech implementation for real estate and construction teams, starting from $4,000/month and cancel anytime. If change orders are where your margin quietly leaks, send us the one project that generates the most change-order friction and we will map it with you.