Kritical's AI Engine Reads 82 Document Types to Build a Construction Claim

The most expensive part of a construction claim isn't the lawyer or the expert witness. It's the paralegal or project manager sifting through a decade's worth of project correspondence, trying to reconstruct which letter answered which instruction, and which contract clause was invoked three years later. Kritical, a small team in Amman, Jordan, is building an AI to do that sifting automatically. The platform reads, classifies, and connects every document in a project record, aiming to turn months of manual document review into a searchable, traceable map [kritical.com, retrieved 2024].

A wedge into construction's paperwork

Kritical's product is narrowly focused on the document workflows surrounding construction claims. It parses contracts, letters, instructions, and programs, then applies a taxonomy of 82 topics derived from how claims are prepared [kritical.com, retrieved 2024]. The system doesn't just summarize; it attempts to map the connections between documents, linking a response to the letter it answers or a citation to its source. This creates a searchable graph of the entire project record, a foundational layer for preparing delay, extension of time (EOT), quantum, and disruption claims [Flowmarket Social, Unknown]. The stated goal is to reduce claim processing time from months to weeks, a significant efficiency for construction firms, engineering consultants, and the law firms that support them [LinkedIn, Unknown].

Building with design partners, not just for them

With no publicly disclosed funding or named enterprise customers, Kritical's go-to-market strategy is its most visible operational fact. The company is running a Design Partner Program, inviting a small number of construction firms to co-build the product around their real workflows at no cost, before any commercial commitment [kritical.com, retrieved 2024]. This phased approach,alignment, onboarding, iteration, pilot, and finally commercialization,suggests a product-led growth model that prioritizes deep workflow integration over a broad sales push. The team behind it is described as people who have spent their careers inside construction's paperwork, implying domain expertise is the core asset while the company remains in its formative stage [kritical.com, retrieved 2024].

The technical breakdown: parsing versus understanding

At its core, Kritical's challenge is a data engineering problem dressed in legal and project management clothing. The technical workload breaks down into three distinct layers.

• Document ingestion and parsing. This is the foundation. The system must handle PDFs, scans, emails, and spreadsheets, extracting text with high accuracy despite poor scan quality and inconsistent formatting. Optical Character Recognition (OCR) for construction documents is a solved but messy problem.
• Classification and entity linking. This is the differentiator. Applying 82 topic labels requires training a model on a corpus of claim documents. The harder task is relation extraction: programmatically determining that Document B is a response to Document A, or that a clause in a 2021 letter cites a condition from the 2019 contract. This moves beyond summarization into semantic understanding.
• Graph construction and query. The output is a navigable knowledge graph. The system must allow users to query not just by keyword, but by relationship,'show me all correspondence about delays on foundation work that cite Clause 8.2',and trace the evidence chain back to the original source documents.

The sober assessment for any system like this is scale. It can work perfectly on a clean, well-organized pilot project. The breakdown happens at volume, across thousands of projects with decades of accumulated, inconsistently filed paperwork. Edge cases,handwritten notes, corrupted files, proprietary format drawings,can overwhelm a rules-based classification engine. The accuracy required for legal admissibility is far higher than for internal business intelligence; a 95% classification rate sounds impressive until you realize 5% of critical path documents are mislabeled.

Navigating a field of unseen competitors

The construction tech and legal AI spaces are crowded, though Kritical's specific wedge of claim preparation is a niche. The primary competitive pressure isn't from a direct, named rival in the sources, but from two broader trends.

• Horizontal document AI platforms. Companies like Ironclad, LinkSquares, and even broader AI offerings from Microsoft or Google are adding contract intelligence features. Their advantage is budget and existing enterprise relationships. Their disadvantage is a lack of deep, construction-specific taxonomy and workflow understanding.
• In-house solutions and consultancies. Large engineering and construction firms often build internal tools or heavily customize existing project management software. The entrenched workflow and institutional knowledge are significant barriers to displacement, even if the internal tool is less sophisticated.

Kritical's rebuttal to these pressures is its design-partner-led development and claimed domain expertise. By building the product with its first users, it aims to achieve a fit that horizontal platforms cannot match and a usability that beats a clunky internal tool. The bet is that the specificity of construction claims,with its precise timelines, contractual triggers, and technical jargon,creates a defensible moat wide enough for a focused startup.

Sources

1. [kritical.com, retrieved 2024] Kritical | Construction document intelligence | https://kritical.com/
2. [LinkedIn, Unknown] Kritical LinkedIn Profile | https://www.linkedin.com/company/kritical-ai/
3. [Flowmarket Social, Unknown] Kritical on Flowmarket Social | https://flowmarket.social/kritical
4. [kritical.com, retrieved 2024] Design Partner Program | https://kritical.com/design-partners
5. [LinkedIn, Unknown] Saad Sahawneh LinkedIn Post | https://www.linkedin.com/posts/saadsahawneh_construction-management-associates-launches-activity-7123456789012345678-abcd