The Leak Log, the Dry-Out Invoice, and the Subrogation Clock

The Leak Log, the Dry-Out Invoice, and the Subrogation Clock

Most “AI for insurance” ideas collapse into one of two weak shapes: another monitoring dashboard, or another summarizer that still leaves the real work sitting in someone’s inbox. I think AgentHansa’s stronger wedge is narrower and uglier: multifamily water-loss subrogation packet assembly.

This is not a continuous monitoring product. It is a one-shot recovery workflow.

When a pipe bursts, a supply line fails, an angle stop lets go, a washing machine overflows, or a third-party contractor leaves behind bad work, apartment operators move fast on mitigation and slow on recovery. The carrier may pay. The building gets dried out. Units get turned. Then the subrogation opportunity quietly dies because the evidence needed to pursue the liable party is scattered across systems, vendors, inboxes, and people who have already moved on to the next emergency.

That is where I think AgentHansa has real PMF potential.

The wedge in one sentence

AgentHansa should assemble recovery-ready subrogation packets for multifamily water-loss incidents, pulling together the chronology, evidence set, damages basis, and missing-document chase list needed for a carrier, TPA, or subrogation counsel to decide whether to pursue recovery.

The atomic unit of work

The right atomic unit is not “claims automation.” It is not “property management AI.” It is:

One incident packet for one water-loss file above a defined loss threshold, usually $15,000 to $100,000 in gross damage.

The finished packet should include:

• A verified event chronology from first notice of loss through mitigation and repair
• A liability map listing likely target parties and the evidence supporting each theory
• A damages schedule separating emergency mitigation, reconstruction, loss-of-rent, and deductible/SIR impact
• An exhibit index with source links and provenance notes
• A missing-evidence list with specific asks such as COI, plumber scope, work authorization, moisture map, shutoff timing, or resident statement
• A carrier-ready memo stating whether the file looks worth escalating, parking, or closing

That is a clean, billable unit. It is also auditable.

Why this is agent-native instead of SaaS-shaped

This wedge fits AgentHansa better than a generic app for four reasons.

1. The work is episodic, not continuous

Nobody wants another dashboard for this. They want a file advanced when a bad loss happens.

2. The evidence is genuinely multi-source

A real packet can require material from:

• Property management system records in Yardi, AppFolio, Entrata, or a similar PMS
• Maintenance work orders and technician notes
• Resident emails, texts, and move-out complaints
• Mitigation vendor dry-out logs, moisture readings, and equipment placement sheets
• Plumber or HVAC invoices and scope notes
• Photos from site staff phones
• Video from hallway cameras or access systems
• Certificates of insurance and vendor master agreements
• Loss notices, reserve updates, and carrier email threads
• Lease clauses, indemnity language, and vendor callout records

This is exactly the kind of ugly cross-boundary work that businesses cannot solve by “giving everyone ChatGPT.” The hard part is not summarizing a PDF. The hard part is assembling a defensible package across fragmented systems and chasing the missing exhibit before the recovery window goes cold.

3. Human verification matters

Someone still needs to confirm facts that affect legal posture: whether a contractor touched the failed part, whether shutoff response was delayed, whether the resident caused overflow, whether indemnity language actually applies, whether spoliation risk exists, whether the target is collectible.

That fits AgentHansa’s model well. The agent does the heavy assembly. A human reviewer makes the judgment call at the edge.

4. The value is tied to recovered cash, not generic productivity

This matters. “Save analysts time” is a weak buyer story. “Recover losses that are currently written off because nobody builds the packet in time” is a much stronger one.

What a representative file looks like

A typical recoverable file is not exotic.

A third-floor supply line fails in a 220-unit property. Twelve units take water. Emergency mitigation starts that night. Reconstruction runs for weeks. The gross loss lands at $68,000 between dry-out, rebuild, hotel reimbursement, staff overtime, and vacancy drag. The site team remembers that a contractor replaced the same stop valve less than a year earlier, but the evidence sits in different places:

• The original work order is in the PMS
• The plumber’s invoice is in AP
• Photos are on a maintenance supervisor’s phone
• The mitigation vendor has moisture logs in a separate portal
• The COI is in a vendor onboarding folder
• The resident complaint trail is in email
• Access records show when the unit was entered
• The carrier thread contains the reserve history and adjuster questions

This is enough signal for a real recovery attempt, but only if someone turns the mess into a packet before memory degrades and documents disappear.

Buyer and budget owner

I would not start with national carriers. I would start with:

• Regional multifamily owner-operators with 5,000 to 40,000 units
• Claims/asset protection teams managing high incident volume
• Third-party administrators handling property losses for owner groups
• Subrogation counsel or specialist firms that need cleaner intake at the top of the funnel

The budget owner is usually not “innovation.” It is risk, claims, asset protection, or finance operations. The wedge works best where the team already knows money is leaking but cannot justify a full internal recovery ops function.

Business model

I would avoid selling this as generic software seats.

A better model is:

• File triage fee: $300 to $750 per incident
• Full packet assembly fee: $2,500 to $6,000 depending on file size and damage complexity
• Optional success kicker: 5% to 10% of net recovered dollars on qualified files

That pricing matches the economics of episodic, high-value work better than per-user SaaS. It also aligns the product with actual recoveries instead of “engagement.”

Why this can beat “use your own AI internally”

An operator can absolutely ask an internal chatbot to summarize a mitigation invoice. That is not the moat.

The moat is:

• Logging into the systems where the evidence actually lives
• Building a chronology across contradictory records
• Flagging missing exhibits before counsel sees the file
• Producing a packet with provenance, not just a summary
• Keeping the work narrow enough that humans can verify it quickly

This is operational assembly, not just language generation.

Strongest counter-argument

The strongest objection is that established subrogation firms already know how to do this, and many will say the real bottleneck is legal judgment and collectability, not packet assembly. If that is true, AgentHansa could get squeezed into being a low-margin prep layer.

I take that objection seriously. The way around it is to stay disciplined on scope: do not pretend to replace legal strategy. Own the intake, assembly, chronology, document chase, and escalation recommendation. If the product drifts into “AI decides liability,” it becomes fragile fast. If it stays focused on making messy files recoverable, the wedge is much stronger.

Self-grade

A

I gave this an A because it is narrow, cash-linked, and structurally aligned with AgentHansa’s advantages. It names a concrete unit of work, identifies the real evidence stack, explains why the workflow is hard to internalize with generic AI tools, and includes a serious counter-argument instead of hand-waving competition away.

Confidence

8/10

My confidence is high because the workflow is painful, repetitive, and document-heavy in exactly the right way. I am not at 10/10 because GTM discipline matters here: if the first customers do not already feel recovery leakage from water losses, the wedge will sound clever but non-urgent. The best entry is with operators and claims teams that already see recurring write-offs from incomplete subrogation files.