For organizations that compete on the strength of their bids, the ability to respond quickly and accurately to complex Requests for Proposal (RFP) is not just an operational priority, it is a competitive advantage. Yet for many organizations, the process of reviewing and interpreting lengthy RFP documents remains largely manual, inconsistent, and time-consuming. The Analytics team at Forvis Mazars recently partnered with a large construction and engineering organization to address exactly this challenge, delivering a proof-of-concept that demonstrates how intelligent systems can transform the way companies engage with unstructured documents.
The Challenge
Our client specializes in the design and delivery of complex precast concrete structures such as parking facilities, commercial buildings, and institutional projects. Securing new work depends on responding to highly detailed RFPs, each accompanied by technical specification documents that can span hundreds or even thousands of pages.
Buried within those documents are the precise requirements that determine project scope, pricing, and risk. Locating and interpreting that information demands significant manual effort from experienced estimators, effort that varies from one individual to the next. Teams have no efficient mechanism to identify or draw upon knowledge from similar projects completed in the past. The result is a bid development process that is slow, labor-intensive, and exposed to avoidable inconsistency.
With a clear ambition to embed artificial intelligence (AI) meaningfully across the organization, Forvis Mazars helped its client identify where an investment in AI could deliver the most immediate and tangible value. Bid development emerged as a natural starting point—a high-volume, high-stakes process where the gap between current practice and what AI could enable was significant.
Our Approach
Rather than committing immediately to a large-scale implementation, the client elected to begin with a focused proof-of-concept. This approach allowed both teams to validate core capabilities, surface practical insights, and establish a clear path forward—while managing cost and risk.
The solution centered on two capabilities: intelligent document processing and similarity-based project matching.
Intelligent Document Processing
Large specification PDFs were ingested into an AI-driven workflow designed to understand not just the content of each document, but its structure. By first analyzing tables of contents and section layouts, the system adapted to variations in format and organization across different RFPs—a critical advantage given the diversity of documents the client regularly encounters.
Well informed document metadata and classification directly improved extraction accuracy. The proof-of-concept solution consistently identified and categorized key requirements, including materials specifications, submission criteria, and compliance obligations. Extracted content was then normalized into structured outputs, transforming previously unmanageable documents into clear, searchable data that estimators could act on quickly. Importantly, this structured data also serves a longer-term purpose: high-quality, consistently organized information is the foundation upon which more autonomous, agentic AI capabilities are built. Every document processed is an investment in future intelligence.
Similarity Analysis & Historical Knowledge
Beyond extraction, the solution addressed one of the client’s most desirable unmet needs: the ability to learn from prior work. Using embedding models, each project’s extracted specifications were converted into similarity representations in vector space. This process enabled the comparison of new RFPs against historical projects based on meaning and context—not just keyword matches.
Estimators could then quickly identify projects with similar characteristics, providing immediate reference points for pricing, scope decisions, and risk assessment. Institutional knowledge that had previously existed only in the minds of individual team members became accessible, consistent, and scalable.
Visualizing the Potential
To bridge the gap between technical capability and business intuition, the proof-of-concept included a lightweight demonstration interface. Rather than aiming for a production-ready tool, it was designed to give stakeholders a tangible sense of what the solution could become, allowing them to review extracted specification summaries in a single view and explore similarity relationships across historical projects. The proof-of-concept approach demonstrated feasibility and value for this particular use case, and built confidence in taking the next step of developing a full-scale, enterprise-grade solution.
The Results
Delivered over approximately eight weeks, the proof-of-concept demonstrated that AI could reliably extract critical specification data from large, unstructured documents; surface relevant historical projects quickly and consistently. As a result, our client recognized that the current manual effort required for bid development could be materially reduced, while improving confidence in pricing and scope decisions.
What It Means for Your Organization
The Forvis Mazars Analytics team helps organizations move beyond isolated AI experiments and build intelligent systems that deliver durable, measurable value. We combine advanced analytics, domain expertise, and thoughtful system architecture to embed AI into end-to-end workflows—improving consistency, unlocking institutional knowledge, and supporting better decisions at scale.
If your organization is navigating complex, document-intensive processes and exploring how AI can create a meaningful competitive advantage, we welcome the opportunity to discuss what a practical path forward might look like. More broadly, organizations that prioritize data quality and structure today are positioning themselves for a future where AI agents can act with far greater autonomy and insight—turning well-governed data into an agentic AI foundation—a strategic asset that compounds in value over time. The question is not whether that future is coming, it is how your organization will prepare for it.