AI & Risk Disclosure

This AI & Risk Disclosure provides transparency about the artificial intelligence capabilities, limitations, and risks specific to the Precon platform. We believe you should clearly understand what our AI can and cannot do before relying on its outputs for professional decisions.

This document is incorporated into and forms part of our Terms of Service.

Precon uses the following AI and automated technologies:

• Large language models (LLMs) — provided by third parties (e.g., Google Gemini) for text analysis, content generation, entity extraction, and classification
• Document processing pipelines — OCR, glyph reconstruction, spatial layout analysis, and pattern matching for construction document parsing
• Natural language processing (NLP) — deterministic rule-based entity extraction (spaCy) combined with LLM-based discovery for novel entity types
• Vector search — semantic similarity search across document chunks for retrieval-augmented generation
• Graph databases — entity relationships and document structure for contextual analysis
• Agentic workflows — multi-step AI workflows with tool use for complex analyses (document analysis, takeoff assistance)

All AI-generated content is probabilistic and may contain errors, omissions, or inaccuracies.

AI Outputs include but are not limited to: document analyses, quantity takeoffs, cost estimates, bid recommendations, scope summaries, contract clause extraction, risk scoring, entity recognition, trade classification, lead scoring, and email intelligence insights.

Hallucination and confabulation. Large language models can generate outputs that appear plausible and authoritative but are factually incorrect — a phenomenon widely known as “hallucination” or “confabulation.” This means the AI may invent quantities, cite nonexistent document sections, fabricate entity names, generate specifications that do not exist in the source material, or present confident answers that have no basis in the underlying data. Hallucinated outputs are indistinguishable from accurate outputs without manual verification against source documents.

Accuracy varies based on the quality and clarity of source documents, the complexity of the construction project, the specificity of the domain, and the inherent limitations of the underlying AI models. Precon implements guardrails (retrieval-augmented generation, source grounding, and confidence thresholds) to reduce hallucination, but these measures cannot eliminate it entirely.

AI Outputs from Precon do not constitute and are not a substitute for:

• Licensed engineering or architectural advice
• Certified professional estimating
• Legal advice from qualified attorneys
• Financial or accounting advice
• Building code compliance determinations or safety assessments

You are ultimately responsible for all professional decisions, actions taken, and failures to act. AI Outputs are decision-support tools that must be validated by qualified professionals.

AI-assisted quantity takeoffs, measurements, material lists, and cost estimates are approximations. They may:

• Miscount or omit quantities from complex drawings
• Misinterpret scale, dimensions, or units of measurement
• Fail to account for field conditions, site-specific variables, or local material pricing
• Produce inconsistent results when processing different versions of the same document

You must verify all quantities against source documents and field conditions before using them as the basis for bids, contracts, or material orders.

AI-generated bid content, scope narratives, and pricing recommendations must be reviewed and approved by qualified estimators and project managers before submission. Precon is not liable for bid errors, omissions, under-bids, over-bids, missed deadlines, or financial losses resulting from reliance on AI Outputs.

Contract clause extraction, risk scoring, playbook compliance assessments, and negotiation suggestions are AI-generated. They:

• May miss clauses or mischaracterize their risk level
• Do not account for the full context of a contractual relationship
• Cannot replace the judgment of a qualified construction attorney
• Do not constitute legal representation or the attorney-client relationship

All contract analysis outputs must be reviewed by qualified legal counsel before being relied upon for contractual decisions.

Insights derived from email scanning are automated interpretations that may:

• Miss important communications or context
• Misclassify the intent or urgency of messages
• Surface irrelevant information as actionable
• Fail to identify time-sensitive opportunities or deadlines
• Incorrectly attribute communications to projects or contacts

Do not rely solely on email intelligence for business-critical decisions. Continue to monitor your email directly for important communications.

AI-generated scopes of work, work breakdown structures, and task narratives are automated drafts based on document analysis and pattern recognition. They may:

• Omit scope items that are implicit in source documents but not explicitly stated
• Include work items that are not applicable to your specific project conditions
• Misattribute scope items to the wrong trade, division, or phase
• Fail to capture exclusions, alternates, or allowances from the source documents
• Use generic language that does not reflect local practices, codes, or conventions

All AI-generated scopes must be reviewed, refined, and approved by qualified project managers or estimators before use in bid proposals, subcontractor agreements, or project execution plans.

Project leads are discovered through automated scanning of publicly available sources. The accuracy of project details, timelines, budgets, and contact information is not guaranteed. Leads may be stale, duplicated, or incorrect. Verify all lead information independently before committing resources.

Automated entity extraction, trade classification, and sheet identification may miss or misidentify entities, trades, quantities, or document types. Accuracy is dependent on source document quality, including scan resolution, OCR clarity, and document formatting consistency.

Dashboards, reports, win-rate calculations, pipeline analytics, and other statistical outputs generated by the Service rely on data you and your team have entered, along with automated inferences. These outputs may:

• Reflect incomplete data if projects, bids, or outcomes have not been fully entered or updated
• Produce misleading trends based on small sample sizes or inconsistent data entry practices
• Display stale information if underlying data sources have changed since the last synchronization
• Aggregate data across time periods, teams, or project types in ways that may obscure important nuances

Analytics are intended to surface patterns and support strategic thinking, not to serve as audited financial or operational reports. Validate critical metrics against your own records before relying on them for business decisions.

Precon relies on third-party AI models and services whose behavior, performance, updates, and availability are outside Precon's direct operational control. Third-party model updates may change output characteristics, accuracy levels, or capabilities without prior notice. Precon monitors for significant changes but cannot guarantee output consistency across model versions.

Customer Data sent to AI providers is processed in accordance with our Data Processing Addendum:

• AI providers do not retain Customer Data beyond the duration of processing a single request
• Neither Precon nor its AI sub-processors train foundation models on Customer Data
• See our Privacy Policy for full data handling details

All critical business decisions must involve material human oversight and professional judgment.

This includes, but is not limited to: bid submissions, contract executions, scope commitments, cost commitments, material orders, subcontractor selections, and any decision with significant financial, legal, or safety implications.

AI Outputs are designed to augment, not replace, professional expertise. They are decision-support tools, not decision-makers.

EU AI Act (Article 50): Precon informs users that content generated through the Service involves AI. We are committed to maintaining AI literacy among our personnel and transparency in our AI systems, consistent with the requirements of Regulation (EU) 2024/1689.

Your obligations: If you operate in a jurisdiction with AI-specific regulations, you are responsible for understanding and complying with your obligations, including any requirements to disclose AI involvement in outputs you share with third parties.

Precon continually works to improve AI accuracy and reliability. We use quality monitoring, user feedback, and advances in AI technology to enhance the Service. If you encounter inaccurate AI Outputs, we encourage you to report them to support@precon.com so we can investigate and improve.

By using the Precon platform, you acknowledge that you have read and understood the risks described in this disclosure. You accept that AI-generated outputs are probabilistic, may contain errors, and require independent verification by qualified professionals before use in professional or commercial contexts.