Uncompromising document review: six analytical dimensions

Package-Driven Checklist Verification

Triage compliance at scale with deep, localized rule matching.

Operational mechanism. Import any standard or custom vertical pack. The system combines local semantic vector comparison with fuzzy text matching to review your document.

Visual delivery. Your document stays untouched on the left; the interactive validation checklist populates on the right. Every discovery maps to its precise bounding box with a dedicated amber highlight.

Structural Data Extraction

Convert raw legal blocks into structured, actionable variables.

Operational mechanism. Automatically isolate corporate counterparty definitions, jurisdiction limits, liability caps, termination milestones and financial markers.

Visual delivery. Generates structured summary matrices ready for export into Excel or Markdown, perfect for initial due-diligence ingestion workflows.

Semantic Redline: compare two versions by meaning

See what actually changed between two drafts, not which characters differ.

Material vs cosmetic, summarised. A local AI model reads both versions and separates the changes that touch rights, obligations, amounts, dates, parties or remedies from pure rewording, writing each one in a single line, for example "Liability cap reduced from €50,000 to €20,000". The material changes rise to the top; the cosmetic ones drop to the bottom.

The net that catches micro-edits. In parallel, a deterministic check compares every number, amount, percentage, date and deadline across the two versions, so "30 to 60 days" or a changed cap is flagged as material even when the surrounding sentence looks identical.

Gone, or just moved? DeepSearch shows where a clause that looks deleted actually went, even when it was merged into another article and rewritten, side by side on the real documents. Confirm or reject each difference and export a tracked report. Runs 100% offline, originals untouched; requires the full 12B AI Engine (16GB+).

Chat with Document

Converse directly with your data room. Demand answers, backed by evidence.

Operational mechanism. Grounded local execution. The chat box answers only using verified context harvested directly from the open file. If an item does not exist, the system issues a definitive "Not Found" rather than hallucinating artificial terms.

Visual delivery. Clicking a conversational citation automatically scrolls the main view, flashes the paragraph and locks a persistent bounding box highlighting the source.

Playbook Review: the clause is there, but is it written the way you want it?

Classify every clause against your firm's negotiating positions, entirely on device.

Operational mechanism. Encode your playbook as three tiers per clause: preferred wording, acceptable fallbacks, unacceptable positions. FolioCore finds each clause and classifies it as Aligned, Fallback or Deviation, with a citation and a one-line rationale grounded in the clause text. Where the engine cannot decide, it says Unclassified, never a guess.

Bring your own template. Point FolioCore at your gold-standard model document (PDF or Word) and it drafts the playbook for you: clauses segmented, checks generated, your standard wording captured as the preferred position. Every draft requires human review before it runs.

What it replaces. The first-pass markup. Each outcome is a negotiation posture: accept, push to fallback, or escalate, with the preferred wording side by side, ready for the redline. Negotiation triage with human confirm/override on every check, not legal advice.

Consistency Check: is the document internally consistent?

One click. No package to choose. The proofreading pass nobody has time for.

Four deterministic engines. Amounts in words vs figures ("one million dollars ($10,000,000)"), including currency mismatches. Broken and ambiguous cross references, resolved against the document's real structure. Defined terms: used but never defined, defined but never used, duplicated definitions. And template placeholders left in: "[INSERT COMPANY NAME]", "{{customer_name}}", blank signature fields.

Trust by design. Pattern and structure analysis first, local AI only to confirm ambiguous candidates: in proofreading, a false positive costs more than a miss. Every finding cites page and paragraph, paired sites shown side by side (definition and usage, words and figures). And when nothing is wrong, FolioCore tells you what it verified: references resolved, amounts parsed, terms examined.