The Broker's AI Advantage: Build Your Data

I spent 15 years in insurance agencies and catastrophe claims. When I heard Matt Fitzpatrick, CEO of Invisible Technologies [Moonshots Podcast], explain that AI is only as reliable as the data it's trained on, one scenario immediately came to mind.

An underwriter staring at 100 properties from a REIT. All they have are ACORDs, supplemental applications, and an Excel sheet to plug into an algorithm.

No context. No visual proof. Just addresses, square footage, roof ages, and COPE data (construction, occupancy, protection, exposure).

OK…some more data sources exist, but as far as they type of data (text inputs only), several dozen photos does not make an app “multi-modal”.

The underwriter can't digest this information well. Neither can the AI systems insurers are rushing to implement. The Excel sheet is just a number crunching math problem setup by actuaries.

The 95% AI Failure Rate Nobody Talks About

Here's what the industry doesn't want you to know: 95% of enterprise AI projects fail to deliver ROI due to inadequate data infrastructure.

MIT research found that 95% of companies see zero measurable bottom-line impact from their AI investments. American enterprises spent an estimated $40 billion on AI systems in 2024.

Most of that money was wasted.

Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data. The AI race is being lost at the data layer, not the algorithm layer.

Insurance and property management face a specific version of this problem. You can't train AI on Excel sheets and PDFs. The data has no visual context, no verifiable proof, no immutable record of what actually exists.

What Happens When You Train AI on PDF Reports

Traditional property condition assessments are subjective by design.

A risk assessment officer walks through a property with a camera. They decide what to photograph. They decide what to write. They decide what might be important in the future.

That's selection bias at every step.

We've relied on human expertise in lieu of total information because it was necessary. A person with a camera and a ladder can't capture every square foot of a 50-building multifamily property.

But now it's not necessary anymore.

360 cameras, drones, digital twins, point clouds, orthomosaics—these tools capture everything. Every square foot, inside and out. No selection criteria. No artistic representation. Just Site Reality.

When AI agents analyze properties for underwriting, claims, or capital deployment decisions, they need to work from the same visual medium that humans use. That's how you get faster, unbiased, accurate assessments.

The Dual-Layer Data Asset You're Building Today

When you, as the trusted commercial insurance broker, start implementing Capture Blocks™ into your workflows today, you're building two things simultaneously.

First, immediate operational value. Underwriters can virtually tour properties in an afternoon instead of scheduling site visits. Your team can answer carrier questions in 30 seconds instead of three days. Property managers can confirm sprinkler systems, count HVAC units, measure roof conditions—all from a digital twin. Clients can leverage actionable intel on their portfolio from any computer.

Second, future AI training infrastructure. You're creating a timestamped, linear visual record of your physical assets. Over time, you'll see degradation patterns. You'll extrapolate economic lifespans. You'll understand when capital needs deployment without your clients sending crews to every roof.

Here's what most people miss: the conversations around the digital twin become training data too.

When you and an underwriter tour a property virtually, discussing roof conditions, sprinkler systems, HVAC concerns that entire transcript feeds into your enterprise AI. The visual capture plus the human knowledge layer creates a corpus of data that generic AI tools can't replicate.

Companies like Drone Deploy and Propeller have millions of digital twins in their servers. But they don't have your conversations. They don't have your enterprise-specific interpretation of what matters for multifamily versus warehouses versus retail.

That's the asset you're building.

The 3-6% Margin That Exists Because of Ambiguity

Every commercial property policy holds 3-6% in reserve for re-inspection and litigation costs.

That margin exists primarily because of information ambiguity. Enter the lawyers!!

When claims happen, the biggest fights aren't about coverage terms. They're about pre-loss condition. What was the roof like before the storm? What was the HVAC condition before the fire? What was the scope of work?

Lawyers build entire practices around these arguments. Florida accounts for 76-79% of the nation's homeowners insurance lawsuits but only 8-9% of claims. Since 2017, over 10 property and casualty companies that offered home insurance in Florida have liquidated, with five liquidating in 2022 alone, with legal costs in 2019 exceeding $3 billion just fighting these lawsuits.

When you have a Capture Block proving pre-loss condition and another proving post-loss condition, you have two sources of absolute truth. You're not negotiating about what was. You're just pricing the delta.

Months of back-and-forth evaporate. Mediation becomes increasingly unnecessary. Everyone works from the same physical dataset.

Carriers can offer credits on policies with Capture Blocks because [at absolute minimum] litigation vectors are minimized. Brokers add value to both policyholders and carrier relationships. Property managers have documentation that speeds claims resolution.

The only people whose business outlooks are diminished: the lawyers.

What Monday Morning Looks Like With Capture Blocks

You're still doing your job the same way. You're still working with carrier partners, reviewing Excel sheets, examining supplementals.

But instead of staring at a PDF and talking about the deal, you screen share and virtually walk the property. You answer questions that make your partners uneasy. You look at pool gate labels, push-to-exit buttons, accessibility features.

You don't have to guess about what you need to verify. You don’t need to continually go back to your clients for the minutiae that makes all the difference.

Can we confirm the property is sprinklered? Jump into a hallway. You can see sprinklers and fire extinguishers in every corridor. Confirmed in 30 seconds.

What's the condition of every HVAC unit? The drone captured all sides, all serial numbers. You can see granular cracking, degradation, maintenance needs.

How many light bulbs are in each hallway? You can count them from the digital twin. Yes…a ridiculous question, but name me a PCA that could even contemplate attempting to answer something so off the wall? YOU CAN.

The Capture Block documents every publicly accessible part of the property. You don't worry about misrepresentation. You can answer exactly what you need to get your markets to sharpen their pencils and get the deal done.

The Competitive Timeline That's Already Started

AI adoption in insurance is accelerating faster than most people realize. Insurers anticipate AI adoption increasing from 14% today to 70% in the next three years.

2026 will be a huge year for enterprise adoption attempts.

The organizations that start building visual datasets today will have years of training data when AI tools mature. Their competitors will struggle with legacy information gaps.

More data over a longer period means more accurate extrapolations. Multiple Capture Blocks across your portfolio, combined with transcripts of every management conversation about those properties, creates an AI that understands what matters to your enterprise.

When you're clients are doing acquisitions, your AI will immediately recognize what's important to their investment philosophy. That only happens if you have the dataset over time. 
WHY…Because they will call you at the beginning of the investment process instead of at the end. Why would you as a broker want to delegate these tools to your carrier partners and continue to be part of black box rating you cannot advocate against?

Training AI on a single snapshot gives you shallow insights. Training AI on years of timestamped visual records plus human analysis gives you competitive advantage.

Why Most Organizations Aren't Doing This Yet

It's not cost. It's not ROI skepticism. It's not inertia.

Organizations simply aren't aware this capability exists.

Brokers and property managers are focused on selling more policies, managing more properties, hitting their numbers. AI feels confusing and distant—something for tech companies, not insurance professionals.

But here's what you need to understand: you don't have to do anything differently than what you're already doing.

You're adding the Capture Block. That expense gets paid for by premium reductions from eliminated litigation reserves. Or policyholders invest in it because they see the value for their onsite management teams. How to structure the money side can be as unique as each client. As the agent, you are in the driver seat on that.

The dataset itself gains value without you actively using it to its highest degree. It exists in the vault, ready to be accessed and assessed when your enterprise is ready to implement AI analytics.

Companies rushing to plug PDFs into AI chatbots will be disheartened by the results. A dozen photos from a property condition assessment two years ago can't revolutionize how you evaluate properties.

That's not how it works.

The Data Infrastructure Decision You're Making Right Now

Andrew Ng, Professor of AI at Stanford, states that if 80% of machine learning work is data preparation, ensuring data quality is the most critical task for any AI team.

You're not preparing for some distant AI future. You're solving today's problems—underwriting ambiguity, claims litigation, capital deployment uncertainty—while building tomorrow's competitive advantage.

The organizations that understand this will dominate their markets. The ones that wait will spend years trying to catch up with incomplete datasets and legacy documentation gaps.

Risk assessment and pricing continue to rely on traditional actuarial models and historical data trapped in static PDFs. Critical information stays difficult to access, resulting in underutilization and oversight.

Over 73% of enterprise data leaders identified "data quality and completeness" as the primary barrier to AI success—ranking it above model accuracy, computing costs, and talent shortages.

The problem isn't the AI. It's the data the AI has to work with.

Property Blockchain™ methodology—delivered through OVRWatch—solves this at the foundation. Immutable, evidence-grade Capture Blocks create the statistically validatable baseline that enterprise AI requires.

You get clarity first. Competitive advantage second. AI training third.

But you have to start building the dataset now.