A delay claim is the most demanding document a project produces: when a project finishes late and the parties dispute who caused it and who pays, the claim package that argues the case is a legal-financial instrument assembled from forensic schedule analysis, time impact analyses, windows analyses, cost build-ups, and a narrative tying causation to consequence, and it is built to recognized standards, the AACE International Recommended Practice 29R-03 for forensic schedule analysis chief among them, because it will be scrutinized by opposing experts, lawyers, and possibly an arbitrator or a court. AI can accelerate the assembly enormously: it can build the time impact analysis fragnets, run the windows analysis, comb the schedule updates for the forensic record, marshal the cost, and draft the narrative, turning months of expert labor into a fraction of the time. But this is a high-stakes document where the causation argument and the expert opinion are owned by the scheduling or claims expert, not the AI, because the claim rests on the expert's professional opinion about what caused the delay, which is a judgment the expert is accountable for and must verify, the schedule analysis, the causation logic, and the factual basis. This lesson designs the delay-claim-package workflow, where AI assembles and accelerates and the human owns the expert opinion the claim rests on.

What a Delay Claim Is and Why It Is the Hardest Document

A delay and disruption claim argues that the project finished later (or cost more) than it should have because of causes the other party is responsible for, and it must prove three things: that the delay occurred (the schedule analysis showing the project slipped), that the other party caused it (the causation argument linking the cause to the delay), and that the delay had a quantifiable consequence (the cost or time the claimant is owed). Each of these is contestable, so the claim package must establish each rigorously: the schedule analysis must withstand an opposing expert's critique, the causation must survive the other side's alternative explanations, and the quantum must be defensible against the other side's accounting, which is why the delay claim is the hardest document a project produces, because it must prove a contested, technical, consequential case to a skeptical, expert audience.

Disruption, distinct from delay, is the loss of productivity caused by the impacts (the crews working less efficiently because of the disruption, even if the project did not finish late), which is harder still to prove because it requires showing the productivity that would have been achieved absent the disruption versus what was achieved, a counterfactual the measured-mile or other recognized methods address but that remains contestable. So the claim package combines the delay case (the schedule slipped, the other party caused it) and the disruption case (the productivity was impaired, the other party caused it), each demanding its own rigorous analysis, narrative, and quantum, assembled into a package that argues the whole. The recognized methods, the AACE RP 29R-03 forensic schedule analysis methods (the windows analysis, the time impact analysis, the as-planned versus as-built, and the others) for the delay, the measured-mile and its kin for the disruption, exist precisely because these cases are contested and the methods give the analysis a defensible, recognized basis, so the claim built to the recognized methods is harder to dismiss than one built ad hoc. The delay and disruption claim is the hardest document because it proves a contested technical case to an expert audience, built to recognized methods that give it a defensible basis.

How AI Assembles and Accelerates the Claim Package

AI accelerates the claim-package assembly at every labor-intensive step. For the forensic schedule analysis, it can comb the schedule updates (often dozens of monthly updates over the project) to build the as-built record, identify the critical-path shifts, and construct the windows for a windows analysis, the tedious schedule-forensics work that consumes expert time. For the time impact analysis, it can build the fragnets (the schedule fragments modeling each delay event's impact) and insert them into the schedule to model the delay's effect, the modeling labor the TIA method requires. For the cost, it can marshal the cost records, build the cost build-up, and tie the costs to the delay events. And for the narrative, it can draft the prose tying the causation to the consequence, presenting the case coherently, the generative drafting at which it excels.

The value is enormous because the claim package is the most labor-intensive document a project produces, often months of expert time, so the AI's acceleration of the schedule forensics, the TIA modeling, the cost marshaling, and the narrative drafting can compress that to a fraction, making it feasible to produce a thorough claim where the labor might otherwise force a thinner one. But the acceleration is of the assembly, the mechanical and laborious parts, not of the expert opinion the claim rests on, which is the crucial distinction: the AI builds the fragnets, runs the windows, marshals the cost, and drafts the narrative, but the expert opinion about what caused the delay, which the schedule analysis supports and the narrative argues, is the expert's professional judgment, not the AI's computation. So the AI assembles and accelerates the package, but the expert owns the opinion the package argues, which means the AI's assembled analysis is the expert's tool, verified and adopted as the basis of the expert's opinion, not an opinion the AI produces. The AI assembles the package fast; the expert owns the opinion the claim rests on.

AI assembles the delay-claim package, building the time impact analysis fragnets, running the windows analysis, combing the schedule updates for the forensic record, marshaling the cost, and drafting the narrative, turning months into a fraction. But the causation argument and the expert opinion the claim rests on are owned by the scheduling or claims expert, not the AI, who must verify the schedule analysis, the causation logic, and the factual basis, because the AI assembles and accelerates and the human owns the expert opinion the claim is built on.

The Expert Owns the Causation Argument and the Opinion

The delay claim rests on an expert opinion: the scheduling or claims expert opines that the delay was caused by the events the claim attributes it to, an opinion grounded in the schedule analysis but ultimately a professional judgment the expert is accountable for, often literally, as the expert may testify to it under oath and be cross-examined on it. This opinion is the heart of the claim, because the delay's cause is contested (the other side argues different or concurrent causes), so the claim turns on the expert's reasoned attribution of the delay to its causes, which is a judgment about causation that the schedule analysis informs but does not mechanically determine, since the same schedule data can support different causation arguments depending on the expert's reasoned analysis of what drove the critical path.

So the causation argument is irreducibly the expert's: the AI can build the schedule analysis (the as-built, the windows, the TIA) that the causation rests on, but the attribution of the delay to its causes, the reasoning that this event drove this critical-path slip and the other side's alternative explanation does not hold, is the expert's professional opinion, which the expert must form, verify, and own, because the expert is accountable for it. The AI can draft a causation narrative, but if the expert adopts it without forming and verifying the causation judgment themselves, the expert is signing an opinion they did not form, which is both professionally improper (the opinion must be the expert's own) and practically dangerous (the expert will be cross-examined on an opinion they cannot defend because they did not reason to it). So the expert owns the causation argument and the opinion, forming and verifying the causation judgment, using the AI's assembled analysis as the basis but owning the reasoned attribution, because the opinion is the expert's professional judgment they are accountable for, which the AI can support but not supply. The expert owns the opinion the claim rests on, the responsible-charge principle at its most consequential, because the expert's opinion is the claim, and an opinion the expert did not form and cannot defend is no opinion at all.

Verifying the Schedule Analysis, the Causation, and the Facts

The expert's ownership requires verifying three things the claim rests on: the schedule analysis (the forensic schedule work the AI assembled), the causation logic (the reasoning attributing the delay to its causes), and the factual basis (the facts the analysis and causation rest on). The schedule analysis must be verified because the AI's assembly can err: a windows analysis with the windows drawn wrong, a TIA with a fragnet modeled incorrectly, an as-built with a critical-path shift misidentified, any of which would make the analysis unsound, so the expert verifies the AI's schedule analysis is correct, the methods applied properly, the critical path correctly traced, the fragnets correctly modeled, because the causation rests on the analysis and an unsound analysis undermines the opinion.

The causation logic must be verified because the attribution of the delay to its causes is the opinion's core, and the AI's drafted causation can be flawed: attributing a delay to a cause the schedule does not actually support, ignoring a concurrent cause, or making a causation leap the analysis does not justify, so the expert verifies the causation logic holds, that the analysis actually supports the attribution and the alternative explanations are addressed, because the causation is the contested heart of the claim. The factual basis must be verified because the analysis and causation rest on facts, the dates, the events, the schedule updates, the cost records, and the AI can draw on these imperfectly, so the expert verifies the facts are accurate, because the claim's whole edifice rests on the factual record and a wrong fact undermines the analysis built on it. So the expert verifies the schedule analysis, the causation logic, and the factual basis, the three foundations of the opinion, because the opinion the expert owns rests on all three, and the expert's verification of each is what makes the opinion sound and defensible. The expert verifies the analysis, the causation, and the facts, the three things the claim rests on, because the expert owns the opinion and the opinion is only as sound as its analysis, causation, and facts.

The Applied Problem: Design the Delay-Claim-Package Workflow

Here is the exercise. Design the delay-claim-package workflow: specify the AI's assembly (the forensic schedule analysis combing the updates and building the windows, the time impact analysis building the fragnets, the cost marshaling, the narrative drafting, to the recognized methods such as AACE RP 29R-03), and the expert's ownership, which separates the expert opinion and causation argument (the expert's professional judgment, owned and formed by the expert) from the assembly (the AI's accelerated labor), with the expert verifying the schedule analysis, the causation logic, and the factual basis. Produce the workflow design that assembles the claim package fast while the expert owns and verifies the opinion the claim rests on.

Produce two things. First, the delay-claim-package workflow design: the AI's assembly and the expert's ownership and verification, in the form that would let a claims team produce a thorough, defensible claim package with the expert owning the opinion. Second, the ownership-and-verification analysis: why the causation argument and the expert opinion are the expert's to own and cannot be supplied by the AI (the responsible-charge principle at its most consequential, because the expert is accountable for the opinion and may testify to it), why the expert must verify the schedule analysis, the causation logic, and the factual basis, and why the AI assembles and accelerates while the human owns the expert opinion, with the reasoning. Pay particular attention to the causation argument, because it is the contested heart of the claim, an opinion the expert must form and defend, which the AI can support with the schedule analysis but cannot supply, since an opinion the expert did not form and cannot defend under cross-examination is no opinion at all.

The deliverable is the delay-claim-package workflow design and the ownership-and-verification analysis, and the lasting product is a designed claim-package workflow that uses AI to assemble the forensic schedule analysis, the TIA, the cost, and the narrative fast, to the recognized methods, while the expert owns and verifies the causation argument and the opinion the claim rests on. This is the capstone of the change-management and claims chapter, where the responsible-charge principle reaches its most consequential application, because the document is a legal-financial instrument argued by an accountable expert, and it applies the responsible-charge ownership, the analysis-causation-facts verification, and the assemble-versus-opine distinction to the highest-stakes document the project produces. The professional who masters this assembles a thorough, defensible claim package with AI while owning the expert opinion the claim rests on, because the opinion is the expert's professional judgment they are accountable for, the causation is the contested heart they must form and defend, and the analysis and facts must be verified, so the AI's assembly accelerates the labor while the expert owns the opinion, which is the only way an AI-assembled claim can be the expert's defensible opinion, because the claim is the expert's opinion and the expert's ownership of it cannot be delegated to the AI that assembled the analysis it rests on.

Selecting the AACE Method Is the Expert's Judgment, Not the AI's Default

Before any analysis runs, someone has to decide which forensic method the claim will use, and that choice is itself an expert judgment the AI cannot make. AACE International Recommended Practice 29R-03 catalogs a family of methods, the windows analysis, the time impact analysis, the as-planned versus as-built, the collapsed as-built, and others, and they are not interchangeable. Each rests on different assumptions about the schedule record, demands different data, and carries different vulnerabilities under cross-examination. A windows analysis needs reliable contemporaneous updates to slice the project into periods; a time impact analysis needs a sound baseline and credible fragnets; a collapsed as-built needs a defensible as-built record and is often attacked for the subtractions it makes. The method is chosen to fit the facts of this project and the quality of its record, and choosing it is the first place the expert's judgment governs the claim.

An AI will happily default to whatever method is most common in its training or easiest to assemble from the available files, and that default may be exactly wrong for the case. If the schedule updates are sparse or unreliable, a windows analysis built on them will be picked apart by the opposing expert, and the claim is weaker than if a method better suited to the thin record had been chosen. The AI can assemble whichever method it is pointed at, fast, but it has no stake in whether that method is defensible for this project, and it cannot weigh the record quality, the contract's analysis requirements, or the tribunal's expectations the way the expert must. Pointing the AI at a method and accepting its output is letting the tool make a decision that belongs to responsible charge.

This is why method selection sits upstream of the verification of analysis, causation, and facts: the expert decides which recognized method the case warrants, directs the AI to assemble it, and then verifies the assembled product. The same false-negative asymmetry that governs the rest of the claim applies here, because a method that quietly does not fit the record is a defect that does not announce itself until the opposing expert exposes it, by which time the claim's credibility is already damaged. The expert who owns the opinion owns the method that produces it, treating the AI's speed as leverage applied to a method the expert chose, never as a substitute for the choice itself.

Key Takeaways

• A delay and disruption claim is the hardest document a project produces: it must prove a contested technical case (the delay occurred, the other party caused it, the delay had a quantifiable consequence) to a skeptical expert audience, built to recognized methods like the AACE RP 29R-03 forensic schedule analysis that give it a defensible basis.
• Disruption (the loss of productivity from the impacts) is distinct from and harder than delay, requiring a counterfactual (the productivity absent the disruption versus what was achieved) that the measured-mile and other recognized methods address but that remains contestable.
• AI accelerates the assembly enormously: combing the schedule updates for the forensic record, building the windows analysis, modeling the TIA fragnets, marshaling the cost, and drafting the narrative, turning months of expert labor into a fraction, making a thorough claim feasible.
• The acceleration is of the assembly (the mechanical, laborious parts), not of the expert opinion the claim rests on: the AI builds the analysis, but the opinion about what caused the delay is the expert's professional judgment, not the AI's computation.
• The expert owns the causation argument and the opinion: the delay's cause is contested, so the claim turns on the expert's reasoned attribution, a judgment the schedule analysis informs but does not mechanically determine, which the expert must form, verify, and own because they are accountable for it and may testify to it.
• An opinion the expert adopts from the AI without forming and verifying it is improper (the opinion must be the expert's own) and dangerous (the expert will be cross-examined on an opinion they cannot defend because they did not reason to it), so the causation is irreducibly the expert's.
• The expert verifies the three things the claim rests on: the schedule analysis (the AI's assembly can err, the windows drawn wrong, a fragnet modeled incorrectly), the causation logic (the attribution must hold and address alternatives), and the factual basis (the dates, events, and records the analysis rests on).
• The artifact: design the delay-claim-package workflow (AI assembling the forensic schedule analysis, TIA, cost, and narrative to the recognized methods; the expert owning and verifying the opinion, causation, analysis, and facts), and analyze why the opinion is the expert's to own (responsible charge at its most consequential) and why the AI assembles while the human opines.