The eleven construction contract clauses that move real money, what the risky version of each one does, what to negotiate for instead, and an honest rating of how reliably software can detect it.
What is contract risk detection, and which clauses does it flag?
Contract risk detection is the automated review of a contract to find and classify the provisions that shift cost, time, or liability onto one party, then rank them by how much money each one can move. In a construction contract it targets a short and well known list: indemnity, liability limits, payment conditions, delay, notice, and flow-down.
That list is short because construction contracts fail in predictable places. The same eleven provisions account for most of the money that moves between parties after a job goes sideways, and they show up in nearly every subcontract regardless of who drafted it. Detection is useful precisely because the target set is stable. You are not asking software to understand your project. You are asking it to find eleven things, tell you what each one says, and tell you which numbers they carry.
The scale is what makes this worth automating. The U.S. Census Bureau estimated the value of construction put in place in May 2026 at a seasonally adjusted annual rate of $2,210.2 billion, in its Monthly Construction Spending release published July 1, 2026. Every dollar of that runs through a contract, and most contracting firms review those contracts under deadline pressure with people who are not lawyers.
Which contract clauses actually cost construction contractors money?
Eleven clauses do most of the damage: indemnity and defense obligations, limitation of liability or its absence, consequential damages waivers, pay-if-paid, no-damage-for-delay, liquidated damages, termination for convenience, retainage, changes and notice-of-claim windows, differing site conditions, and flow-down by reference. The matrix below covers each one.
Read the last column honestly. Detection reliability is not a marketing number, it is a statement about what the text of a contract can and cannot tell you. A provision that is either present or absent, and that carries a number, is easy to detect. A provision whose real effect depends on a state statute, a document you were never sent, or a judgment about your own exposure is not.
| Clause | What the risky version does | What to push for | Detection reliability |
|---|---|---|---|
| Indemnity and defense | Requires you to defend and indemnify the upstream party for claims arising from the work, including claims caused by that party's own negligence, with no dollar cap and a duty to defend that starts on allegation rather than on a finding of fault | Indemnity limited to your own negligence, defense triggered by adjudicated fault rather than by allegation, and a cap tied to your insurance limits | High for presence and for sole-negligence language. Moderate on effect, because several states void broad forms by statute and the governing law decides |
| Limitation of liability | No cap at all, so your maximum exposure is unbounded and unrelated to your contract value. This is the risk most often found by its absence rather than its wording | An aggregate cap stated as a dollar figure or a percentage of contract value, with carve-outs narrowed to willful misconduct | High. Absence is easy to test for and a cap is a number. Low on whether the cap is adequate, which is a business judgment about your exposure |
| Consequential damages waiver | Missing entirely, or one-directional so you waive and the other party does not, leaving you exposed to lost profits, lost use, and financing costs that dwarf your fee | A mutual waiver, with any carve-outs listed explicitly and kept narrow | High for presence and mutuality. Moderate on whether the carve-outs are broad enough to swallow the waiver |
| Pay-if-paid | Makes the owner's payment a condition precedent to your payment, so if the owner never pays you may never be owed, regardless of whose fault the nonpayment was | Pay-when-paid with a stated outside date, or a clause that shifts only the timing and not the risk of owner insolvency | Moderate to high for detecting condition precedent language. Low on enforceability, which turns on the governing state and often on case law |
| No-damage-for-delay | Limits your remedy for delay to a time extension only, even when the delay was caused by the other party, so you absorb extended general conditions and acceleration costs | Compensable delay for delays within the other party's control, with time extension as the remedy only for genuinely neutral events | High for presence. Moderate on enforceability, which differs between public and private work and between states |
| Liquidated damages | A per-day rate set far above any real forecast of the owner's loss, with no cap on total accrual and no matching remedy running the other way | A daily rate tied to a documented forecast, an aggregate cap, and a mutual bonus or shared savings where the schedule matters that much | High for presence and for the daily rate. Low on whether the rate is an unenforceable penalty, which is fact specific |
| Termination for convenience | Lets the other party terminate at will while limiting your recovery to costs incurred, with no profit on work performed, no demobilization costs, and no compensation for the margin you priced into the remaining work | Recovery of costs incurred, profit on work actually performed, demobilization and restocking costs, and a defined settlement window | High for presence. Moderate on whether the compensation formula is adequate, since the gap is usually in what the clause omits |
| Retainage | A high percentage held on the full contract value through final completion, with release conditioned on events you do not control, such as the owner accepting the entire project | A reduced percentage after substantial completion, release of your retainage on your scope rather than the whole job, and a securities substitution option where state law allows | High for the percentage and the stated release trigger. Moderate on whether the terms comply with the applicable state retainage cap |
| Changes and notice of claim | Very short written notice windows, often measured in days, with an express waiver of the claim if notice is late. These clauses kill more valid claims than pricing disputes do | Notice windows measured in business days from actual knowledge, no automatic forfeiture for a late notice that caused no prejudice, and a clear list of who counts as notified | High for extracting the day count. Low as a standalone control, because a flag does not calendar the deadline for you |
| Differing site conditions | Absent entirely, or paired with disclaimers elsewhere in the contract stating that you accept all subsurface risk and that the geotechnical reports are for information only | An express differing site conditions clause covering both materially different subsurface conditions and unknown unusual conditions, with an equitable adjustment | High for presence or absence. Moderate on effect, since a disclaimer buried in a different article can undercut it |
| Flow-down by reference | One sentence binding you to every obligation the prime owes the owner, incorporating a prime contract, general conditions, or specifications you were never given a copy of | The flow-down list stated explicitly, the incorporated documents attached, and a stated order of precedence when the documents conflict | Low to moderate. The incorporating sentence is easy to flag, but software cannot read a document that is not in the file. The correct output is a request for the missing document |
What does a risky indemnity clause look like, and what should you push for?
The risky version requires you to defend and indemnify the upstream party for claims arising out of your work even when that party caused the loss, with no dollar cap and a duty to defend that triggers the moment a claim is alleged. The defense obligation is usually the more expensive half, because it starts before anyone has determined fault and it does not wait for a finding.
Several states have limited how far this can go, which is why the same clause behaves differently on two jobs in two states. Texas enacted a broad anti-indemnity statute in 2011, effective January 1, 2012, that voids most fault-shifting indemnity in construction contracts:
"Except as provided by Section 151.103, a provision in a construction contract, or in an agreement collateral to or affecting a construction contract, is void and unenforceable as against public policy to the extent that it requires an indemnitor to indemnify, hold harmless, or defend a party, including a third party, against a claim caused by the negligence or fault, the breach or violation of a statute, ordinance, governmental regulation, standard, or rule, or the breach of contract of the indemnitee, its agent or employee, or any third party under the control or supervision of the indemnitee, other than the indemnitor or its agent, employee, or subcontractor of any tier."
Texas Insurance Code § 151.102
California reaches a narrower version of the same result. Civil Code § 2782(a), subject to exceptions it cross-references, makes void and unenforceable construction contract provisions purporting to indemnify the promisee against liability arising from “the sole negligence or willful misconduct of the promisee.” Note how much narrower that is. California voids indemnity for the promisee’s sole negligence, while the Texas statute reaches claims caused by the indemnitee’s negligence or fault more generally, subject to its own exceptions including an employee-injury carve-out at § 151.103.
What to push for: indemnity limited to claims caused by your own negligence, a defense obligation that attaches to adjudicated fault rather than to a bare allegation, and a cap tied to the insurance limits you actually carry. Detection will reliably tell you the clause is there and whether it reaches the other party’s own negligence. It will not tell you whether your governing state voids it.
What happens when a contract has no limitation of liability, and why does a consequential damages waiver matter?
With no limitation of liability clause, your maximum exposure under the contract is unbounded and has no relationship to what you are being paid. This is the single most valuable finding automated review produces, because it is a finding of absence, and absence is exactly what a human reader skims past.
The consequential damages waiver is the companion problem. Direct damages are the cost to fix the work. Consequential damages are the owner’s lost profits, lost use of the facility, extended financing costs, and lost rent, and on a commercial project those can exceed your entire contract value by an order of magnitude. Private nonresidential construction ran at a seasonally adjusted annual rate of $738.7 billion in May 2026 according to the Census Bureau’s Monthly Construction Spending release, and that is the work where a downstream contractor’s fee is smallest relative to the owner’s business loss if the building does not open on time.
Two failure patterns show up constantly. The first is no waiver at all. The second is a one-directional waiver, where you give up consequential damages against the other party and they keep theirs against you. The second is worse, because it reads like protection.
What to push for: a mutual waiver of consequential damages, an aggregate liability cap stated as a dollar figure or a percentage of the contract value, and carve-outs from the cap kept narrow and specific rather than a general exception for anything the drafter later decides was serious. Detection catches presence, absence, mutuality, and the cap figure with high accuracy. Whether the cap is large enough for your actual exposure is a judgment about your business, not a property of the text.
Is a pay-if-paid clause enforceable, and how is it different from pay-when-paid?
Pay-when-paid changes when you get paid. Pay-if-paid changes whether you get paid at all, by making the owner’s payment a condition precedent to the obligation running to you. Enforceability varies by state, and in several states the answer comes from case law rather than a statute, so this is the clause where a general rule is most likely to be wrong on your job.
The language itself is usually unmistakable once you know what to look for. The California Supreme Court reproduced the operative wording from the subcontracts before it in a 1997 decision: “Receipt of funds by Contractor from Owner is a condition precedent to the Contractor’s obligation to pay Subcontractor under this Agreement, regardless of the reason for Owner’s nonpayment, whether attributable to the fault of the Owner, Contractor, Subcontractor or due to any other cause.” The phrase to search for is condition precedent. Where it is missing and the language is ambiguous, courts frequently read the clause as pay-when-paid, which preserves the obligation and only delays it.
New York’s highest court reached the question first:
"We hold that a pay-when-paid-provision which forces the subcontractor to assume the risk that the owner will fail to pay the general contractor is void and unenforceable as contrary to public policy set forth in the Lien Law § 34."
West-Fair Electric Contractors v. Aetna Casualty & Surety Co., 87 N.Y.2d 148 (N.Y. 1995)
California followed in Wm. R. Clarke Corp. v. Safeco Ins. Co. of America, 15 Cal. 4th 882 (1997), concluding that “pay if paid provisions like the one at issue here are contrary to the public policy of this state and therefore unenforceable because they effect an impermissible indirect waiver or forfeiture of the subcontractors’ constitutionally protected mechanic’s lien rights in the event of nonpayment by the owner.” Both courts got there through lien law rather than through payment law, which is why the reasoning does not travel automatically to states with different lien statutes.
Other states legislate it directly. North Carolina provides that “Payment by the owner to a contractor is not a condition precedent for payment to a subcontractor and payment by a contractor to a subcontractor is not a condition precedent for payment to any other subcontractor, and an agreement to the contrary is unenforceable” at N.C. Gen. Stat. § 22C-2. Virginia Code § 11-4.6(B)(2) bars the same thing in private construction contracts, with a narrow exception where the paying party is insolvent or a debtor in bankruptcy, and requires the same payment, notice, and interest terms to run down to lower-tier subcontracts on projects other than single-family residential worth more than $500,000.
The risk pay-if-paid transfers is counterparty failure, and that risk is not theoretical. According to the Bureau of Labor Statistics Business Employment Dynamics data on establishment survival, 56.5 percent of construction establishments that opened in the year ended March 2020 were still operating five years later in March 2025. Signing a true pay-if-paid clause means accepting the chance that the party above you is in the other group.
What to push for: pay-when-paid with a stated outside date, so the obligation survives even if payment upstream stalls. Detection flags the condition precedent language with good accuracy. It cannot tell you whether your governing law will enforce it, and it should not pretend to.
Can a no-damage-for-delay clause stop you from recovering delay costs?
Often yes on private work, and considerably less often on public work, because several states limit these clauses by statute on public projects. The clause caps your remedy for delay at a time extension, which means you keep working while absorbing extended general conditions, idle equipment, and acceleration costs.
Public construction spending ran at a seasonally adjusted annual rate of $541.2 billion in May 2026 per the Census Bureau, which is the segment where the statutory limits actually apply. California addresses it in Public Contract Code § 7102, which provides that clauses limiting a public agency’s liability to a time extension “for delay for which the contractee is responsible and which delay is unreasonable under the circumstances involved, and not within the contemplation of the parties, shall not be construed to preclude the recovery of damages by the contractor or subcontractor.” Virginia Code § 2.2-4335(A) voids any public construction contract provision purporting to waive, release, or extinguish a contractor’s right to recover for unreasonable delay caused by the public body.
Read the exceptions carefully, because they matter more than the rule. Both statutes expressly preserve provisions requiring notice of delay, providing for liquidated damages, and establishing dispute resolution procedures. Your right to recover survives. Your obligation to give notice on time does not go away, and a blown notice window will end the claim before anyone reaches the question of whether the no-damage-for-delay clause was enforceable.
What to push for: compensable delay for delays within the other party’s control, with a time extension as the sole remedy reserved for genuinely neutral events like unusually severe weather.
When do liquidated damages and termination for convenience become the expensive clauses?
Liquidated damages become expensive when the daily rate is untethered from any real forecast of loss and nothing caps the total. Termination for convenience becomes expensive when the compensation formula leaves out profit on work you already performed and the demobilization costs you will actually incur.
The federal construction clause shows what a disciplined version looks like. FAR 52.211-12 sets liquidated damages as a stated dollar amount “for each calendar day of delay until the work is completed or accepted,” and adds that if the government terminates the contractor’s right to proceed, “liquidated damages will continue to accrue until the work is completed” and are in addition to excess costs of repurchase. The structure is clear: one number, one trigger, one accrual rule. What makes a private liquidated damages clause dangerous is usually not the concept but the absence of an aggregate cap.
On termination, the federal clause is again the useful reference point. FAR 52.249-2 states that “The Government may terminate performance of work under this contract in whole or, from time to time, in part if the Contracting Officer determines that a termination is in the Government’s interest,” and gives the contractor one year from the effective date of termination to submit a final settlement proposal, extendable in writing by the contracting officer. Private termination-for-convenience clauses often keep the at-will right and drop the settlement machinery.
What to push for: a documented basis for the daily liquidated damages rate and an aggregate cap on accrual, and on termination, recovery of costs incurred plus profit on work actually performed plus demobilization, with a defined window for submitting the settlement. Detection reads the daily rate and the presence of a cap with high accuracy. Whether a rate is an unenforceable penalty is fact specific and not something a document reader can decide.
Why do short notice and claim windows kill more claims than bad pricing?
Because a valid claim submitted late is often no claim at all, and the windows are short enough that they expire during normal project chaos. This is the highest-value, lowest-glamour finding on the page: the day counts buried in the changes and claims articles decide more money than the pricing terms everyone negotiates hard.
The federal clauses illustrate the shape of the problem. Under FAR 52.243-4, a contractor “must assert its right to an adjustment under this clause within 30 days” after receiving a written change order or furnishing written notice of a constructive change. Separately, no adjustment for a constructive change “shall be made for any costs incurred more than 20 days before the Contractor gives written notice as required.” Two clocks, one for asserting the right and one that quietly deletes costs you incurred before you spoke up. The FAR prescription that accompanies the clause adds that the 30-day period “may be varied according to agency procedures,” so even the federal number is not universal.
Differing site conditions works the same way. FAR 52.236-2 requires written notice promptly and before the conditions are disturbed, and provides no adjustment after final payment under the contract. There is no day count at all, which is worse in practice than a short one, because “promptly” is decided later by someone reviewing what you did.
The outer limit is generous and misleading. Under the Contract Disputes Act at 41 U.S.C. § 7103, each claim must be submitted within six years after the claim accrues, with certification required for claims over $100,000. Six years sounds like comfort. It is not, because the contract’s own notice provision will have extinguished the claim long before the statutory window closes. Private subcontracts frequently run shorter than the federal numbers.
What to push for: notice windows measured in business days running from actual knowledge rather than from an event you may not have witnessed, no automatic forfeiture where a late notice caused no prejudice, and a named recipient so notice cannot fail on a technicality. Detection extracts these day counts reliably. It does not put them on anyone’s calendar, and a flag that nobody diaries is not a control. See construction change order management for how the change and notice cycle runs in practice.
What does retainage language actually control, and what should you negotiate?
Retainage language controls three things: the percentage held, what it is held against, and the event that releases it. The third one is where the money hides, because a clause that conditions your release on the owner accepting the entire project ties your cash to work you do not perform and cannot control.
The federal approach is a useful contrast to standard industry subcontract forms, which commonly hold a flat percentage on everything through final completion. Under FAR 52.232-5(e), retainage on a fixed-price construction contract is not automatic. If the contracting officer finds satisfactory progress was achieved, payment is authorized in full, and only if satisfactory progress has not been made may the contracting officer “retain a maximum of 10 percent of the amount of the payment until satisfactory progress is achieved.” Retainage there is a remedy for poor performance with a stated ceiling, not a default withholding.
What to push for: a reduced percentage after substantial completion, release measured against your own scope rather than the entire project, and substitution of securities where state law permits it. State retainage statutes cap the percentage and set release timing in many jurisdictions, and those caps override contrary contract language where they apply. Detection reads the percentage and the stated release trigger accurately. Whether that percentage is legal on your project is a state-law question.
What does flow-down by reference obligate you to, and why is it the hardest clause to detect?
Flow-down by reference binds you to obligations contained in documents that are not in front of you, usually through a single sentence stating that you assume toward the contractor all obligations the contractor assumes toward the owner. It is the hardest clause to detect meaningfully because the risk is not in the sentence, it is in the document the sentence points at.
Federal work makes the scale visible. FAR 52.244-6, in the version dated October 2025, requires the contractor to insert a list of specific clauses into subcontracts for commercial products and commercial services, and that list runs to 24 items numbered (i) through (xxiv), several of which are themselves conditional on dollar thresholds or on flow-down being required by the referenced clause. The same clause adds that a contractor “may flow down to subcontracts for commercial products or commercial services a minimal number of additional clauses necessary to satisfy its contractual obligations,” which is permissive rather than mandatory. So even on federal work, where the required set is published, whether a given clause belongs in your subcontract depends on facts outside the subcontract itself.
On private work there is no published list. The incorporated document might be a prime contract, a set of general conditions, project specifications, an owner’s manual of procedures, or all of them, and the order of precedence among them may be undefined.
What to push for: the flow-down obligations listed explicitly in your subcontract, copies of every incorporated document attached before signing, and a stated order of precedence when documents conflict. The honest detection outcome here is not a risk score. It is a flag that says a document referenced in this contract is not in the file, followed by a request to produce it. That is a workflow problem more than an analysis problem, which is why it belongs in a system that tracks documents rather than in a one-time review.
How reliable is automated clause detection, and where does it fail?
Automated detection is reliable at three things and unreliable at a fourth. It reliably finds whether a clause is present, reliably extracts the numbers a clause carries, and reliably identifies language patterns like condition precedent or sole negligence. It is not reliable at telling you what a clause will do to you, because that depends on facts the document does not contain.
The failure modes are specific and worth naming. Detection cannot resolve conflicts between a clause and a disclaimer sitting in a different article, which is how differing site conditions protection gets quietly undone. It cannot read a document that was never attached, which is the flow-down problem. It cannot apply the governing state’s law to decide enforceability, which is the pay-if-paid, indemnity, no-damage-for-delay, and retainage problem all at once. And it cannot judge whether a liability cap is adequate, because adequacy is a fact about your balance sheet and your insurance, not about the contract.
There is also a quieter failure worth flagging: detection reports risk, and risk reports do not change outcomes on their own. A short notice window that gets flagged and then never enters anyone’s calendar produces the same result as a notice window nobody read. The value of detection is realized when the flag becomes an obligation someone owns with a date attached. See AI contract administration for construction for how flags become tracked obligations, and AI document analysis in construction for how extraction works on a single document.
What does this page not cover, and when do you need a lawyer?
This page does not give legal advice, does not resolve which state’s law governs your contract, and does not tell you whether any specific clause is enforceable on your project. Clause enforceability varies by state, by project type, by whether the work is public or private, and by delivery method, and several of the rules described here come from court decisions rather than statutes, which means they can change without a legislature acting.
Bring in a construction lawyer licensed in the governing state when any of the following is true. The contract contains a pay-if-paid provision and you are not certain of the governing law. The indemnity clause reaches the other party’s own negligence. There is no limitation of liability and the project value is material to your firm. You are considering signing a flow-down by reference without having read the incorporated documents. Or a claim is already live and a notice window is running, in which case the deadline matters more than the analysis.
Talk to your contracting officer, not a vendor, on federal work when the question is what a FAR clause requires on your specific contract. The clause text is public, but its application to your facts is theirs to determine in the first instance.
What this page is for: knowing which eleven provisions to look at first, what the risky version of each one does, what to ask for instead, and where automated review genuinely helps rather than where it is claimed to. See best construction contract management software for how firms evaluate the tooling category, and redlines inside the contract for how Contracts Connected handles a vendor’s changes.
How does Contracts Connected handle clause risk?
Contracts Connected reads a contract, extracts the provisions on this list, and records what each one says and which numbers it carries, so the eleven clauses that decide the money are identified before signature rather than discovered during a dispute. Where a contract incorporates a document by reference that is not in the file, the result is a request for that document, not a confident score computed without it.
The part that matters more than extraction is what happens next. A flagged notice window becomes a tracked obligation with a date and an owner. A missing prime contract becomes an open item. Our co-founder ran a large federal general contracting firm for 15 years, about 100,000 documents a year.
Related reading
- Construction change order management
- AI contract administration for construction
- AI document analysis in construction
- Best construction contract management software
- Payment bond requirements by state
- Subcontractor compliance automation
Frequently asked questions
Which construction contract clauses does automated risk detection flag most reliably?
Automated review is most reliable on clauses that are either present or absent and that carry a number: liquidated damages rates, retainage percentages, liability caps, and notice day counts. It is very good at catching a missing limitation of liability clause, because absence is easy to test for. It is weakest on flow-down by reference, where the risky language points to a document the reviewer may never have received.
Is a pay-if-paid clause enforceable in construction contracts?
It depends entirely on the governing state, and in several states the answer comes from case law rather than a statute. New York and California courts have both held true pay-if-paid provisions unenforceable as against public policy. Virginia and North Carolina bar them by statute. Other states enforce them if the condition precedent language is clear. Never assume the clause is dead or alive without checking the governing law.
What is the difference between pay-if-paid and pay-when-paid?
A pay-when-paid clause is about timing: it delays payment until the general contractor is paid, but the obligation to pay eventually survives. A pay-if-paid clause is about risk: it makes the owner’s payment a condition precedent, so if the owner never pays, the subcontractor may never be owed. Courts often read ambiguous language as pay-when-paid, so the difference usually turns on whether the words condition precedent appear.
Can a no-damage-for-delay clause actually be enforced?
Often yes on private work, and much less often on public work. Several states limit these clauses by statute on public projects. California voids provisions that cap the public agency’s liability to a time extension for unreasonable delay the agency caused, and Virginia voids provisions waiving recovery for unreasonable delay caused by the public body. Both states preserve notice requirements and liquidated damages clauses, so the notice obligation still binds you.
What is the most dangerous clause that automated contract review is worst at catching?
Flow-down by reference. A single sentence binding you to every obligation the prime owes the owner can incorporate hundreds of pages you were never sent. Software can flag the incorporating sentence with high accuracy, but it cannot read a document that is not in the file. The detection result you want is not a risk score, it is a request for the missing prime contract.
Does automated contract risk detection replace a construction lawyer?
No. Detection tells you which provisions are present, what they say, and which numbers they carry. It does not tell you whether a clause is enforceable in your state, whether a liability cap is adequate for your actual exposure, or whether a liquidated damages rate is a penalty. Those are legal and business judgments. Use detection to shorten the list a lawyer reviews, not to skip the lawyer.