AI for construction claims, delay analysis and contract risk in 2026: what each tool actually produces

Contract risk: before you sign and when something changes

First Rule, Document Crunch and Provision all flag risky clauses and cite the language behind each flag. First Rule's assistant, Marten, checks a prime contract or subcontract against construction playbooks with risk levels, preferred language and fallback positions, and its pages name pay-if-paid, indemnity, liquidated damages and notice requirements. Its top tier lists flow-down risk analysis between prime contract and subcontract. Document Crunch's CrunchAI surfaces risk such as flow-down clauses, and its Project Assist covers notices, RFIs and submittals during the job. Provision's Risk Review adds severity levels, default-position comparison and RFI and redline drafts, and reads contracts together with drawings, specs, RFIs and addenda.

Our reading, for claims work: the most useful output is not the risk score but a cited list of notice clauses, time bars and change procedures that the field team can act on in week one. Ask each vendor to show exactly that on a contract you already know well.

Forensic delay analysis: the after-the-fact account

Steelray Delay Analyzer is the only product we read that names a forensic method: it says it runs an automated half-step analysis with the Daily Progress Method per AACE RP 29R-03 (MIP 3.4) on Primavera P6 XER updates, quantifies concurrent and but-for delays, and produces a contemporaneous critical path and longest-path change analysis for each window. It says delay experts and expert witnesses use it (vendor claim). It does not describe itself as AI.

SmartPM lists windows analysis, delay and compression metrics and, on its Controls plan, time impact analysis, connecting to existing CPM schedules. A customer quote on its site says windows analysis went from weeks to minutes or hours (customer claim). It does not describe its product as AI either. Our reading: on the delay side the established tools are deterministic schedule math, and that is arguably what you want in front of an arbitrator.

AI project controls: early warning while the job is live

Nodes & Links lets users ask a schedule what changed since the last update and which changes drove delay on or near the critical path, logs who changed what and when, and says every answer is auditable back to source. Foresight lists schedule quality and change tracking, Quantitative Schedule Risk Analysis (QSRA) and completion forecasting, and claims forecasts 2x more accurate than bottom-up schedules (vendor claim). nPlan forecasts activity-level uncertainty from what it says are 750,000+ historical schedules and names highways, rail and heavy infrastructure. ALICE simulates scenarios to de-risk and recover schedules from P6 or Microsoft Project files.

These are prospective tools. Our reading: their value in a claim is indirect. They can surface a slip early enough to give timely notice and to document mitigation, and their change logs can preserve what each update showed. None of the four says its forecast is a forensic delay analysis.

The cost record: change orders and T&M

Clearstory (formerly Extracker) manages T&M tags with signatures, change order request logs shared between GC, subs and owners, rate libraries and backcharges. Its AI agents draft CORs and T&M tags from voice notes, photos and takeoffs, review incoming CORs against a checklist and approved rates (closed beta), and assign a reason category to every COR. The vendor says rate libraries in heavy civil can exceed 10,000 line items and that it does not train AI models on customer data.

Our reading: a signed, dated, rate-matched T&M tag is often the strongest quantum evidence a contractor has. Ask how edits after signature are tracked and how the log exports.

Pricing

Five of the ten publish prices. Steelray Delay Analyzer: $3,990 per user per year or $390 per user per month. SmartPM: Essentials $12K per year for up to 50 projects, Controls $25K per year. First Rule: Apprentice $795 per month, Proficient $1,495 per month (listed as available mid-2026), Mastery custom. Provision: Risk Review from $100 per project, Chat Agent from $50. Clearstory: Basic free (5 T&M tags and 5 CORs per month); paid tiers through sales. Nodes & Links, Foresight, nPlan, Document Crunch and ALICE do not list prices (Document Crunch offers a free trial; ALICE quotes by project size).

What none of them claims

This is our reading, not a vendor claim: no page we read says its output has been accepted as expert evidence in a specific arbitration or court, and none says it replaces a scheduling expert, a cost expert or counsel. Steelray comes closest by naming an AACE recommended practice and saying experts use it. Treat every tool here as a faster way to assemble and check evidence that qualified people still evaluate.

What to test before relying on any of them

These are our suggestions, not findings. For delay tools, rerun a closed project with a known expert result and compare delay days per window, the critical path in each window, and the concurrency findings. For contract tools, use a contract with known notice and time-bar traps and count what each tool catches or misses. For any tool, confirm you can export the inputs, the outputs and a record of every update, so the work can be reproduced by the other side's expert.