01
Define what the company cares about
Set the fields, document types, diligence checks, synonyms, and output expectations every property workspace then follows.
Field dictionaryPackage templateSchema
CRE Underwriting Intelligence Workspace
Underwrite CRE deals from a single cited source of truth.
Define the underwriting facts, documents, and diligence standards your company cares about. Fync AI turns each property package into structured deal data, flags gaps, enriches the file with outside sources, and helps teams review, ask, and generate from verified evidence.
Document extraction Company field dictionary Missing-data detection External data enrichment Review queues Cited chat & credit memo drafts
Sources
Offering Memo
Rent Roll
Appraisal
Census FEMA BLS Vendor
Evidence Engine
OCRExtractCross-checkCite
Property Knowledge Base
Company-defined · cited
Decisions
Review Queues
Cited AI Chat
Credit Memo
evidence in · decisions out
The problem
The deal file is fragmented before underwriting even starts.
Analysts retype numbers from PDFs, reviewers chase source pages, and outside research lives in browser tabs instead of the deal file. Teams need more than storage. They need a controlled evidence workflow.
Raw borrower PDFs
Offering memos, rent rolls, appraisals, and operating statements arrive as inconsistent PDFs.
Spreadsheet copies
Analysts retype key numbers into spreadsheets with no durable citation back to the source.
External data lookups
Flood, demographics, assessments, and market data get checked in a dozen separate tabs.
Untracked decisions
What was approved, by whom, and why lives in inboxes and tribal memory.
scattered inputs
One cited source of truth
The product
Define the standard once. Apply it to every property.
Fync AI starts with your underwriting model: the fields, documents, required checks, synonyms, and output expectations your team cares about. Every property workspace then applies that standard to the package it receives.
01 · You define
Your underwriting standard
Field dictionary, required documents, and internal output templates: the model your team already underwrites by.
02 · Fync fills
A cited Knowledge Base
Documents, public data, and vendor reports are extracted, enriched, and cross-checked into evidence-backed facts.
03 · Your team uses
Confident decisions
Resolve exceptions in review, ask cited questions, and draft credit memos from reviewed facts.
End to end
From intake to credit memo without losing the source.
Borrower intake
· one connected line ·
committee-ready memo
Feature pillars
Nine connected capabilities, one evidence model.
01 · Company standards
Define what your underwriting team cares about.
Companies define the facts, documents, synonyms, sections, and output requirements they care about. The system uses that structure to classify documents, extract fields, detect missing data, route review, answer chat, and generate internal documents.
Workspace field dictionary: labels, keys, data types, units, synonyms
Required fields and deal-package requirements
Schema candidates when a document holds an unmapped label
Field dictionary
fin.noi currency required
val.cap_rate percent required
physical.units number optional
amenities.ev_charging candidate
Submitted
Rent Roll
Present Fetched
HUD FMR
Failed Ordered
Appraisal
Processing Submitted
Borrower PFS
Missing Ordered
ESA Phase I
Blocking Output
Credit Memo
Ready 02 · Deal package
Find out what is missing immediately.
The deal package turns a messy intake into a readiness map. Borrower uploads, fetched data, ordered reports, generated analyses, and final outputs roll into one view, so the team can see what is present, missing, blocked, or ready to use.
Borrower submittedMachine fetchedThird-party orderedInternally generatedFinal output
03 · Extraction
Turn borrower PDFs into structured underwriting facts.
Upload package PDFs once. Fync AI classifies each document, parses pages and tables, extracts facts into your field dictionary, and verifies every value against its source page. Reprocessing never overwrites approved facts.
PDF-only · 50 MB limit · 39 document types · user-correctable classification
PDF Classify OCR + Parse
Pages + Tables Extract Verify cite
Knowledge Base + schema candidates
In-place NOI fin.noi
$2,940,000
Citation verified · 0.96
Awaiting approval
Operating Statement · p.62 Open PDF
Approve Reject Edit
04 · Knowledge Base
Every property fact is referenced and cross-checkable.
The Knowledge Base organizes company-defined facts by section, shows value status, source, page, and last update, and opens every fact into an evidence drawer with citation, approval, alternate values, and audit trail. Citation verification stays separate from human approval.
18 sections · search & filter · PDF / API / manual / computed / generated evidence
05 · Review queues
Keep humans in control of the numbers that matter.
Fourteen review queues turn machine output into a controlled workflow. Reviewers see what needs attention, open the supporting evidence, choose a resolution, and write a decision reason into the audit trail.
conflicts · missing critical · unverified citations · stale data · entity resolution · +9
NOI conflict
$2.94M (appraisal) vs $3.01M (offering memo)
Use currentDifferent basisReject both
Missing Appraisal
required diligence document
FRED macro skipped
app-level key not configured
Census GeocoderFEMA FloodBLS UnemploymentProperty AssessmentCounty Permits HUD FMR · keyFRED · key CoStar · vendorTitle · vendorPhase I ESA · vendor
implemented requires key vendor scaffold
06 · External data
Research the deal from outside sources automatically.
Fync AI plans, runs, retries, and audits external data adapters. Public sources and vendor scaffolds attach to the deal file the same way uploaded documents do, so research stops living in browser tabs. Absent provider keys are app-level configuration, not company blockers.
See the data provider network →07 · Cited chat
Ask the property Knowledge Base. Get answers with citations.
Chat answers from structured property data first, then falls back to document evidence when needed. Every citation opens into the same evidence drawer used across the workspace.
structured data first · document evidence fallback · every answer cited
What supports the NOI?
Answered from approved facts
In-place NOI is $2,940,000 on a trailing-12 basis, supported by the operating statement and corroborated by the appraisal.
Operating Statement · p.12
Appraisal · p.41
Credit Memo Ready to generate
Depends on
Term SheetRent RollOperating StatementAppraisal
Generate Credit Memo
08 · Generated outputs
Draft credit memos from approved, citation-verified facts.
Generate credit memo drafts from approved, citation-verified facts. Additional output types are dependency-gated, so teams can see what is ready, blocked, or still missing before any work product is created.
credit memo drafts live today · 7 more output types dependency-gated
09 · Operations
Built for auditable pipelines, not black-box demos.
The Developer view exposes pipeline state without leaking secrets. Operators can inspect dictionary configuration, document artifacts, extraction runs, vector chunks, enrichment runs, and environment presence booleans.
artifacts · runs · vector index · env presence (no secret values)
developer · run_a1f9c2 · env live
r2://parcel/p1/h1/ ├─ source.pdf ├─ parsed/markdown.md ├─ parsed/tables.json └─ facts.json (40 facts)
run_a1f9c2parsed
run_b73e10needs_review
enrich_hudskipped_requires_key
Data provider network
Connect the sources that shape a CRE underwriting decision.
Borrower documents, public data, and vendor reports flow into one evidence engine, then out to review, chat, and generated work product.
Property Knowledge Base
evidence engine
Borrower package
Offering MemoRent RollOperating StmtAppraisal
Term sheet
Sources & Uses · Sponsor PFS
Public data
CensusFEMABLSPermits
Vendor / ordered
CoStarTitlePCAPhase I
Key-gated
HUD FMR · FRED · Census ACS
Decisions
ReviewChatCredit Memo
cited facts → work product
Use cases
What underwriting teams do with Fync AI.
02
Screen a new acquisition package
Upload the OM, rent roll, operating statement, term sheet, appraisal, and sponsor materials and watch readiness fill in.
UploadExtractionPackage
03
Prepare for quote committee
Open package blockers, missing critical fields, and unapproved facts, resolve exceptions, and draft a memo outline.
Missing dataReviewCredit memo
04
Replace manual fact checking
Open any fact and see source document, page, excerpt, citation match, approval state, alternates, and audit trail.
EvidenceCitationsAudit
05
Bring public data into the deal file
Run Census, FEMA, BLS, assessment, permits, HUD, and FRED adapters; results become cited facts alongside PDF extractions.
EnrichmentAPI evidenceFreshness
06
Ask a cited question
“What supports the NOI?” or “Show the package blockers.” Answers route to facts, package state, metrics, or document search.
Chat routesCitationsVector search
07
Standardize the underwriting data model
Map extracted labels to a canonical field model; unmapped labels become schema candidates instead of disappearing.
SynonymsSchema candidatesSections
Trust & control
AI where it helps. Human approval where it matters.
The system separates citation verification from reviewer approval. A fact can be extracted and source-matched without being approved. Reviewers approve, reject, edit, resolve conflicts, and log reasons directly against the evidence. Every decision is captured in the audit trail.
Machine
Citation verified
Human
Awaiting approval
Reviewer actions are logged with who, when, prior value, and reason.
Approve Reject Edit
Platform
Built to be inspected, not taken on faith.
Fync AI is built around persisted artifacts, typed workflows, audit trails, encrypted storage, and workspace-level isolation. Every document run, enrichment, fact, review action, and generated output is traceable.
Elastic scale
Spikes in uploads and extraction scale out automatically. No capacity planning.
Encrypted & isolated
Documents and facts are encrypted in transit and at rest, isolated per workspace.
Always-on reliability
Managed, redundant infrastructure keeps deal files available when you need them.
Auditable by default
Every fact, run, and review decision is persisted and traceable end to end.
SOC 2-aligned practicesEncryption in transit & at restRole-based accessFull audit logging
Product surface today
18
Knowledge Base sections
14
Review queues
19
Enrichment adapters
39
Document types
8
Generated output types
5
Package source categories
FAQ
Answers for underwriting teams evaluating Fync AI.
What is Fync AI?
Fync AI is a CRE underwriting intelligence workspace. Companies define their underwriting standard once — fields, required documents, synonyms, and outputs — and Fync AI applies it to every property package. Every fact is extracted from documents or external data with a verifiable citation back to the source.
What does Fync AI do that storage and spreadsheets can't?
Fync AI is a controlled evidence workflow, not a file store. It classifies documents, extracts fields into your dictionary, runs public and vendor enrichment, detects gaps and conflicts, routes exceptions to reviewers, and generates credit memo drafts — all anchored to citations and an audit trail.
What document types does Fync AI support?
Fync AI supports 39 document types across borrower-submitted PDFs (offering memos, rent rolls, operating statements, appraisals, term sheets, sponsor PFS), third-party reports (appraisal, Phase I ESA, PCA, title), and internally generated artifacts. PDF uploads are capped at 50 MB and document classification is user-correctable.
How does the citation system work?
Every extracted fact stores its source document, page, excerpt, and citation confidence. Citation verification is performed by the system; reviewer approval is a separate human action. A fact can be source-matched without being approved, and reviewers can approve, reject, edit, or resolve conflicts directly against the evidence.
Which external data sources are connected?
Public-data adapters cover Census Geocoder, FEMA Flood, BLS Unemployment, property assessment, and county permits. HUD FMR, FRED, and Census ACS are app-level key-gated. Vendor scaffolds exist for CoStar, title, Phase I ESA, and PCA. Nineteen enrichment adapters in total.
What outputs can Fync AI generate?
Credit memo drafts are generated today from approved, citation-verified facts. Seven additional generated output types are dependency-gated so teams can see what is ready, blocked, or still missing before any work product is created.
How is the platform secured?
Workspace-level isolation, encryption in transit and at rest, role-based access, SOC 2-aligned practices, and full audit logging. The developer view exposes pipeline state — artifacts, runs, vector chunks, and env presence — without leaking secret values.
Build a more reviewable underwriting workflow.
Give your team one place to intake packages, resolve blockers, verify source evidence, and draft credit materials from reviewed deal data.