loop
Email Book a call →

Investor brief · Confidential · 2026

The pipeline every profession
already runs —
finally getting faster.

Book a 30-minute call → Take a peek
$0B+ lost annually to manual data gathering across knowledge intensive industries — not to fraud, not to overhead — to copy-pasting.

What LOOP is in one line

LOOP is an ai toolkit that helps professionals automate the pipeline from documents to decisions — and gets smarter every time your team uses it. Every action becomes a reusable workflow — without writing a line of code.

Time per decision4.0 hrs→0.7 hrs
Team cost / decision$192→$34
Improvement per decisionZero→Compounds
Live in private lending Self-funded Not currently raising
scroll

01The pipeline

Other AI extracts data. LOOP captures judgment.

Every action your team takes — pull, correct, flag — is teaching the system how the work actually gets done. Across every stage from documents to decisions, the expertise that lives in your senior reviewer's head becomes the system itself.

2B+

documents uploaded to team directories every year across lending, insurance, legal, and finance. Almost none of them talk to each other.

1.8 hrs

every single day spent by knowledge workers searching and gathering information — per employee, across every workflow. McKinsey Global Institute

01 · Inputs Docs Any document, any format, any source.
Bank statements Pro formas Rent rolls Appraisals
02 · Extract Data Structured extraction with citations.
Value + confidence Source citation Anomaly flags Correction logging
03 · Apply Logic Your existing model — imported once.
Excel model import Formula preservation Cell-to-doc mapping Live recalculation
04 · Output Decisions Verified, cited, exportable output.
DSCR · LTV · NOI Every value cited Credit memo export Full audit trail

In practice — five steps the team takes

01 Open the deal Documents from your existing storage, organized as a deal — the way your team thinks about work.
02 See the raw inventory Every value, table, and labeled number across all documents — surfaced unsorted, uninterpreted.
03 Pull what matters Drag the values that go into your model. Each pull is recorded as an extraction step.
04 Correct & flag Override what's wrong, flag what concerns. Each correction tunes the prompt; each flag adds a rule.
05 Approve Sign off on the analysis. The system has been recording every step in order.
Reads from your existing document stores SharePoint Google Drive Box OneDrive Dropbox + API No data migration required.

LOOP connects any document to any decision model — and turns every correction into a self-improving competitive advantage.

02The core insight

The first software your team writes by working.

LOOP is not an AI decision tool. It's a container that gets configured through use. Three or four saved processes in, the institutional knowledge is encoded inside LOOP, impossible to extract. The correction log is the moat. Compounding by design.

After the first deal — what the system asks

“Save this as a process?”
The system has been watching. Every action your team took becomes a reusable workflow that runs automatically on the next deal — and sharpens with every correction. Your team just programmed the system without writing a line of code.
1

Expert reviews extraction

Extracted values surfaced with source citations. Accept, correct, or flag each field — inline, in seconds.

2

Every action logged automatically

Accept, correct, flag, dismiss — each recorded with source document, value delta, timestamp.

3

Corrections improve extraction

Prompt logic updated immediately. Fine-tuning dataset accumulates for future model training.

4

Accuracy rises, trust compounds

Fewer corrections per decision. Switching cost grows with every decision processed.

“The correction log is institutional memory made portable — a mirror that gets sharper every time the team uses it. It exists only inside organizations that use LOOP, and grows more valuable with every decision processed.”

Personal library Workflows your team has saved. Compounds within one customer. After three or four saved processes, switching cost is the institutional knowledge encoded in the library — impossible to export, impossible to recreate elsewhere.
Shared library Workflows that worked, surfaced as templates. Compounds across all customers. Processes that have been run across enough decisions with low correction rates are auto-surfaced for new entrants. New underwriters don't start from scratch — they inherit institutional knowledge from everyone who came before.

Every interaction is a signal

PULLED → tells LOOP which fields matter for this lender
CORRECTED → tells LOOP where extraction failed and how to do better
IGNORED → tells LOOP what to deprioritize in the inventory
FLAGGED → tells LOOP what your lender considers risk
What LOOP built

Five primitives. That's it.

  • A document reader
  • A review surface
  • An action recorder
  • A process runner
  • A correction logger
What the customer builds

Everything else — without writing code.

  • The extraction rules — by what they pull
  • The calculations — by what they compute
  • The flag conditions — by what they flag
  • The workflow sequence — by the order they work
  • The entire decision logic — by doing the job

LOOP doesn't know what a good pro forma looks like. Doesn't know your DSCR threshold. Doesn't know your credit policy. The customer brings all of it. LOOP just builds the surface that captures it — which is exactly why big tech can't copy it.

The non-interference boundary — the structural reason this stays defensible.

03Vertical strategy

Fifteen verticals. One architecture. The order is the strategy. Redeploying, not rebuilding.

Same pipeline across every regulated profession. The mismatch between decision volume and ease of adoption is the strategy: start with small markets reachable in weeks, earn the right to large markets that take years.

Hover or tap a vertical for detail
Live · Design partner Near-term · Year 1 Year 2 Year 3 / Future
Live · Design partner Private Lending Bank stmts · pro formas · rent rolls Loan approval / decline
Next · Year 1 Construction Lending Draw requests · inspection · budgets Draw approval
Next · Year 1 SBA & Community Banking Business financials · tax returns · apps Credit approval
Next · Year 1 Equipment Finance Appraisals · maintenance · depreciation Advance rate
Year 2 CRE Brokerage CoStar comps · leases · appraisals Price / list / pass
Year 2 Insurance Underwriting Inspection reports · loss histories Bind / price / reject
Year 2 Legal Due Diligence Contracts · reps · warranties Risk flags / sign-off
Year 3 PE & Asset Management CIMs · financials · market data Pass / proceed
Year 3 Healthcare Prior Auth Clinical notes · diagnosis codes Approve / deny

Plus six more verticals identified — Factoring & Invoice, Corporate Functions, Government & Regulatory, Commercial Banking Credit, Venture Debt, Family Office. See all 15 by volume or by ease of adoption.

Why private lending first

The wild-west wedge: high pain, no incumbents, fragmented buyers, and document infrastructure that hasn't been touched by software. Sales cycles in weeks, not quarters. The flywheel can compound here in real-world conditions long before a larger vertical's procurement cycle even opens.

Roadmap

A second document intake layer is planned for external data sources — property records, county assessors, CoStar, Crexi, comparable sales platforms. Private lenders pull this data manually for every deal. Automating external fetch addresses their three largest stated pain points: doc-chasing, data entry, and data spreading.

04Market opportunity · Two frames

Top-down anchors to labor cost. Bottom-up anchors to software spend. Both bracket the real opportunity.

TAM · Top-down $200B+

Knowledge worker time on document-to-decision workflows globally. ~25M target professionals at $80–200K loaded cost. Recapture 20% → $1T+ economic value → SaaS captures a fraction.

TAM · Bottom-up $50B

15 identified verticals × operators × realistic seat count × price point. IDP market projected $49.7B by 2035 at 32% CAGR (Spherical Insights, 2025).

SAM · Serviceable $12B

US-based regulated professional services across LOOP's five near-term verticals. ~20,000 target firms × $500–$2,000/mo per firm.

SOM · Obtainable $290M

Private lending, construction lending, and SBA in the US. ~25,000 target lenders × 10% penetration × $200/seat/mo × 2 seats avg, over 3 years.

05Competitive landscape

Tools that touch one stage. LOOP is the only one wired across all four.

LOOP is not priced as a replacement for any single tool below. It replaces the combination — plus the hours of manual work that sit between them.

For scale: DocuSign automates one stage of this pipeline and is worth $9B. LOOP automates all four.

Tool What it does Price Gap LOOP closes
OcrolusDocument processing / bank stmt analysis$500–2,000/moExtraction only — no logic, no verification UI, no flywheel
BloomaCRE deal analysis platform$1,000–3,000/moSingle vertical, no model import, no doc-to-cell mapping
nCinoBank operating system / loan origination$50,000+/yrEnterprise sales, 12+ month implementation
DocusignE-signature and document workflow$25–40/user/moSignature only — no extraction, logic, or decision output
ABBYY VantageIntelligent document processing (IDP)$3,000–15,000+/moHorizontal IDP — no domain flywheel, no model import
Microsoft CopilotAI assistant embedded in Office$30/user/moHorizontal — no vertical depth, no domain decision logic
LOOPDocs → Data → Logic → Decisions pipeline$299–1,799/moFull pipeline — extraction, model import, verification UI, correction flywheel

06Pricing

Priced against team time, not software.

A mid-level team member runs $48/hr fully loaded. LOOP recaptures 3.3 hours per decision. At ten decisions a month that is $1,584 of monthly labor value recaptured. The Starter tier costs $299. Every tier delivers a minimum 4× ROI.

“Businesses hire 5 people. Only 4 show up to work.”

American knowledge workers spend 1.8 hours every day searching and gathering information — the equivalent of one full work-week per month, lost to manual document handling.

— McKinsey Global Institute

Beyond the time saved, five things become possible that simply weren't before.

01 Consistent process across every deal Same workflow, every decision — consistency at scale, without hiring more reviewers.
02 Auto-flagged deviations When a deal doesn't fit the process, the system surfaces it — not because someone checked.
03 Refine without losing prior work Every saved process versions forward. v2 incorporates what was learned across v1's deals.
04 Compare deals against each other Same extraction structure across every deal exposes portfolio-level patterns that were invisible before.
05 Hand work to a junior — safely Attach a process; trust they're working inside the senior's framework, not reinventing it from scratch.
Starter
Up to 15 decisions/mo · 1 seat
$299/mo
$3,588 / yr
  • Document extraction
  • Process builder
  • SharePoint sync
  • Personal process library
Recommended
Professional
Up to 50 decisions/mo · 3 seats
$699/mo
$8,388 / yr
  • Everything in Starter
  • Excel model import
  • Document-to-cell mapping
  • Shared process library
Enterprise
Unlimited seats · Unlimited decisions
$1,799/mo
$21,588 / yr
  • Everything in Professional
  • External data integrations
  • White-label option
  • API access · Custom templates

Run the math

ROI calculator

Move the sliders to model your shop. The math mirrors the live deployment: 4.0 hrs → 0.7 hrs per decision, 3.3 hrs recaptured.

Defaults reflect the memo benchmarks. Adjust to your team's reality.
Hours recaptured / month82.5
Labor value recaptured$3,960
LOOP monthly cost$699
Net monthly value$3,261
Monthly ROI
vs LOOP cost at chosen tier
5.7×
Per-decision time savings, hourly rate, and tier are independently adjustable. Annualized values are simply ×12.

07Revenue scenarios

Three scenarios. Same pricing. The variable is churn.

All three share the same tiers and addressable market. What differs is adoption speed, churn rate, and tier mix. The single variable with the largest impact is monthly churn — at 1% the correction flywheel has done its job; customers cannot leave without abandoning institutional knowledge encoded over dozens of decisions.

The flywheel is the churn defense. At 1% monthly churn, customers can't leave without abandoning institutional knowledge encoded over dozens of decisions. At 3.5% the product is useful but not yet sticky. Every correction logged moves LOOP from the left side of that spectrum toward the right.

All ARR figures computed live from new customers, tier mix, churn rate, and expansion. Adjust the toggle above to see how the simulation responds.

08Traction

Live, not deck-ware.

LOOP Active Deals dashboard at the live design partner deployment: 30 deals in pipeline, extraction accuracy rising from 62% on deal 1 to 88% on deal 30
The flywheel, working. Phoenix-based private lender, today: 30 active deals processed. Extraction accuracy compounding from 62% on deal one to 88% on deal thirty. The chart that unlocks every subsequent funding stage is already moving — no longer a future trigger.

Live deployment with a private lending operation in Phoenix, AZ — 30 active deals in pipeline, design partner paying and engaged.

Correction logging instrumented — every accept, correct, flag, dismiss captured with source document, value delta, and timestamp. The flywheel is running.

SharePoint integration complete — real deal folders ingested and classified, Sites.Read.All permission scope authenticated.

Excel underwriting model import — existing lender models connected without rebuilding, formula preservation, cell-to-doc mapping.

Extraction tested against 30 real deals — retail, multifamily, mixed-use, non-standardized borrower documents across all types.

Full-stack application live at colonial-os.com — React + FastAPI + Postgres on Hetzner, white-labeled for the design partner.

Interactive wireframe at loop-mocks.netlify.app — clickable surface showing the document → data → logic → decision flow.

09Why now

The window is open. But not for long.

Frontier AI models crossed the extraction quality threshold in 2024 — reliable enough to be useful on day one, correctable enough to compound continuously. The window to build a proprietary self-improving correction dataset before a well-funded incumbent enters this space is open. The pain point is too large for it to stay open for long.

The technology is ready

Frontier models can now extract structured data from non-standardized documents with enough accuracy to be useful immediately. Two years ago this wasn't true.

The incumbents aren't watching

nCino serves banks. Ocrolus serves mortgage. No one is purpose-building for the private lending workflow with a compounding data moat. The category is uncontested.

10Funding strategy

Bootstrap until the flywheel proves itself.

Plan: bootstrap through the first five external customers. One paying customer proves the product works, not that strangers will buy it. Every month spent building the correction dataset deepens the flywheel and strengthens the negotiating position for any future raise. The trigger for the first round is a single chart: extraction accuracy on decision one versus decision fifty.

Now — Year 1 Bootstrap No raise Trigger: Design partner live and paying. First external customer signed. Build the correction dataset. Prove repeatability beyond relationship sales.
Year 1–2 Pre-Seed $500K–1.5M · $5–8M cap Trigger: 5–10 paying customers across 2+ verticals. Accuracy chart decision 1 → 50. First hire. Sales motion. Second vertical fully ramped.
Year 2–3 Seed $2–4M · $12–20M Trigger: 10–20 customers. Organic referrals. NRR above 110%. Sales team. Second vertical scale. Begin fine-tuning proprietary model on correction data.
Year 3–4 Series A $8–15M · $40–80M Trigger: ARR $1–2M. NRR above 120%. Third vertical entering pipeline. Platform scale. Third and fourth vertical. Enterprise motion. External data at scale.

The chart that unlocks everything

Extraction accuracy on decision one versus extraction accuracy on decision fifty. If that number moves materially, the correction flywheel is empirically real. That single chart — not the memo, not the financial model — converts a curious investor into a convinced one. Building the instrumentation to capture it starts on day one.

Next steps

Get involved — investors and early adopters.

For prospective customers

A working demo against live deal packages.

If your team underwrites commercial real estate, construction, or SBA loans and spends significant time on document gathering and data entry, LOOP was built for you. Design partners receive preferred pricing, direct influence over the roadmap, and early access to the process library as it develops.

Request a demo → Email

For investors

LOOP is not currently raising.

The decision to raise will follow proof of the feedback flywheel in a live multi-lender environment. Conversations with aligned investors are welcome now — long before any term sheet. The pre-seed trigger is a single chart: extraction accuracy on decision one versus decision fifty.

Book a call → Email

LOOP is the infrastructure layer that converts document-heavy workflows into automated decision pipelines that compound over time.