The question a board member always asks isn't "what does the document say." It's "how do I know you're right — and which document wins if they disagree?" I built the answer to that question into the system itself.
HOA and condominium governing documents are not written to be understood. They are written to be legally defensible — which means they are dense, cross-referential, layered across multiple document types, amended over decades, and frequently contradictory when a newer rule conflicts with an older declaration.
Every time a board member, association manager, or homeowner asks "can we do this?" — the real question underneath it is a chain of harder questions: What does the declaration say? What do the bylaws say? Do they agree? Are there amendments? Which version of the rules applies? If two documents say different things, which one has higher legal authority?
I managed this problem manually for 14 years. I know exactly how many board meetings go sideways because someone challenges an interpretation, or because no one can immediately cite the exact section that supports the decision being made. The friction is expensive, it erodes board confidence, and it slows every governance decision it touches.
The core insight driving BoardPath's architecture is that document quality and document authority are two separate problems that both have to be solved before you can trust an answer.
Stage 1 — Extraction Pipeline (The Quality Problem)
A governing document corpus is not a clean PDF corpus. It includes scans from the 1980s, handwritten amendments, faxed notices, and documents that have been photocopied so many times the text is barely legible. A single extraction model fails on this material.
BoardPath uses a three-stage OCR pipeline: LlamaParse handles primary extraction for clean and semi-clean documents. Low-confidence outputs route to a MistralAI-powered redundancy layer for enhanced extraction. Documents that still fall below the 93% confidence threshold route to Google Vision as a tertiary fallback — the strongest available tool for worst-quality scan material. Every document in the corpus gets the extraction quality it requires, not a one-size-fits-all pass.
Stage 2 — Hierarchical Document Weighting (The Authority Problem)
Not all governing documents carry equal legal weight. A declaration supersedes bylaws. Bylaws supersede rules and regulations. Amendments modify — and in some cases override — the documents they amend, but only for the specific provisions they address.
BoardPath assigns authority rank to every document type in a client's corpus upon ingestion. When an answer draws from multiple sources, the system knows which document's language controls — and surfaces that explicitly in the citation layer, not just in the answer text.
The hardest design problem in BoardPath wasn't extraction quality or retrieval accuracy. It was trust. Specifically: how do you get a board member who has been skeptical of technology their entire adult life to act on an AI-generated answer?
The answer I arrived at was counterintuitive: don't hide the uncertainty. Surface it. Show exactly how the answer was constructed, how confident the system is in each element of it, and why. A layperson shouldn't need to understand AI to evaluate the output — they should be able to read a plain-language scorecard and make their own judgment.
Every BoardPath answer is accompanied by a confidence score evaluated across five dimensions: authority rank of the cited source, directness of the citation (explicit vs. interpreted), ambiguity flags where language is open to interpretation, conflict detection where multiple documents address the same issue differently, and state statute compliance risk where the answer touches areas of HOA law. The score isn't a single number — it's a legible breakdown. A board member can see exactly what went into it.
This design decision also drove the corpus completeness check. If a client's corpus is missing documents — a common occurrence with older associations — the system flags that the answer may be incomplete because relevant governing language may exist in documents that weren't ingested. Confidence in the answer is bounded by confidence in the corpus.
BoardPath is currently in the final stage of pre-launch preparation, with a live sandboxed public demo in active development and pre-seed investor conversations underway. The system handles the full governance Q&A workflow: document ingestion, extraction, hierarchy weighting, answer generation, confidence scoring, conflict detection, amendment inheritance resolution, and professional DOCX correspondence generation.
The Transparent Confidence™ scoring layer is the product's primary differentiator — and the feature I'm most proud of, because it solves a human problem that has nothing to do with AI capability and everything to do with AI trustworthiness.