Reviews change orders against the contract, schedule, and procurement record. Flags scope already in base contract, markup above your agreed cap, schedule impacts already paid via T&M, and unit-price math errors — with dollar impact per finding.
Six categories cover the bulk of where CO reviews go wrong in the field. Every flag carries a severity tag, the dollar impact, and the source citation — the contract clause, schedule activity, or prior PO line that triggered it.
Sub claims $42K to relocate three plumbing risers. The plumbing scope in the executed contract specifies "all risers per drawing P-203" — which includes the relocated positions. We cross-reference the CO description against contract scope language and base drawings.
Sub bills 14 days of delay impact at $3,200/day for the same period they billed 92 hours of T&M crew time. Both can't be true. We pull T&M logs from the procurement record and surface the date overlap.
Sub's CO carries a 15% O&P markup. The executed subcontract caps CO markup at 10%. We surface the contract clause, compute the corrected total, and flag the $4,260 delta on a $28,400 CO.
180 LF × $43.50/LF = $7,830. The CO says $8,140. A $310 transcription error stacked across 14 lines becomes $4,200. We recompute every unit-price line and flag deltas above your rounding tolerance.
Master agreement: journeyman electrician $86/hr loaded. This CO bills $94. Could be a year-end escalation (valid) or rate creep (not). Flagged with the master rate, the CO rate, and the rate-escalation clause to verify against.
CO traces back to RFI-047 (slab edge condition at column C-7). RFI was answered as "clarification of existing scope" — not a design change. We flag the entitlement gap with the RFI response language for the PM to confirm.
One screen. Left side: the CO exactly as the sub submitted it. Right side: every finding the engine surfaced — severity-tagged, dollar-tagged, citation-linked back to the source.
A reviewer's brief, formatted for the next CO meeting. Findings ordered by dollar impact. Each finding deep-links to the source (contract clause, schedule activity, prior pay app). PM signs off, denies, or routes for negotiation — the audit trail captures the decision.
A CO review tool that only reads the CO will catch math errors. The real value is in the cross-checks — the contract clauses, the schedule activities, the prior procurement record. We pull from each, then surface what doesn't line up.
Pulls the executed subcontract's markup cap, escalation clause, and rate ceilings — checks every CO against those constraints automatically.
When a CO claims schedule impact, we cross-check the underlying activities — were they actually on the critical path during the claimed period? Were they delayed by an owner-directed change?
Pulls pay-app and T&M logs for the affected sub and period. Surfaces overlaps between CO-claimed work and previously billed work — the most common margin leak in our pilots.
This is what we mean by project memory. Each engine on its own catches what it can read. The chain catches what no single document reveals.
Every reviewed CO lands in a single project register — statused, dated, dollar-impact tracked. Owner-side meetings get a clean export. Internal meetings get the full register including disputed-and-denied COs with reasoning.
Status, CO number, vendor, brief description, current dollar value, days impact, approval state. Disputed amounts surfaced; specific sub-vendor margins and internal markup workings hidden. The view your owner-rep can take to the OAC.
Everything above, plus the engine's findings per CO, your PM's decision, dollar deltas, and source citations. The audit trail you keep in your project file in case of dispute.
Optional: auto-generate the CO section of your monthly owner report from the register. Pulls the period's approvals, current pipeline, and forecast impact into a one-page narrative the owner can read in 90 seconds.
A $24M GMP project averages 38 change orders over its life. Pilot data on engine reviews against unreviewed baselines:
Transcription mistakes, unit-price miscalcs. Caught manually by good PMs maybe 30% of the time. Caught by the engine: ~100%.
Where markup exceeds the agreed cap by 1–5 points. Easy money. Most firms either notice or don't depending on who's reviewing that week.
The big one. Sub adds work that was already in their base scope. Hardest to catch because it requires re-reading the original subcontract — which most PMs don't have time for.
On a portfolio of 6–10 active jobs. The number scales with portfolio size; the engine's value scales with it linearly.
Reading the contract, cross-checking the schedule, pulling T&M logs, computing the math. The engine does the first pass in 45 seconds; the PM verifies the findings in 4–6 minutes.
Project meetings spend less time arguing about whether a CO is valid — because the reviewer's brief already shows the work. Owners report higher confidence in the CO log over time.
Numbers above are typical for our pilots and not guarantees. Recovery varies by project size, subcontractor mix, and how aggressively your firm currently challenges COs.
CO Review reads cleanly on its own, but the cross-checks — the part that actually saves money — require the other engines feeding it.
AIA G701, AIA G701/CMa, ConsensusDocs 717, plus any PCO submitted as PDF, Excel, or Word. Custom proprietary forms get auto-mapped to a common schema. If your sub sends it as a JPEG of a hand-marked PDF, we OCR and parse — might surface a confidence flag, but it'll work.
No — this is a review tool, not an approval router. The output is a reviewer's brief. Your PM, project executive, or owner approves the CO through whatever flow you use today (Procore, manual signatures, e-signature platform). Most firms run the engine as a checkpoint before the formal approval step.
Pulls the executed subcontract from the project's documents, parses the scope language plus referenced drawings, and maps CO descriptions against that scope. Confidence drops when the base scope is itself ambiguously written — flagged "verify with PM" rather than firmly denied.
Yes. Same flow, perspective flipped. The owner's CO to the GC gets reviewed against the GMP contract, schedule float ownership, and any markup the prime contract allows. Useful when you're pushing a CO upstream to the owner and want to make sure your own paperwork doesn't have soft spots.
Yes — per-firm and per-project. When you override a "scope already in base" finding because the original scope language was actually ambiguous, the engine records the override and the reasoning. Future COs against the same scope language get flagged with a lower-confidence label, with the prior override surfaced. Over a portfolio, the engine learns what your firm actually treats as "in scope" vs "out of scope" beyond the literal contract text.
Live demo with your real subcontract and three or four real PCOs from a recent job. You keep the output. Disputed dollars from a working project usually pay for the call several times over.
Start free — 14 days, no card →