Itsharkz
Schedule a Call
For Insurers and Brokers

LiquiAI: AI Agent for Insurance Claims Handling

Analysis and recommendation ready before the claims handler opens the case

Book a demo on a real claim
RODO / GDPR

What Is LiquiAI

A claims handler spends most of the time reading, not deciding. Every claim comes with anywhere from a few to a dozen or more documents: the claim form, invoices, repair estimates, photos, medical records, correspondence. LiquiAI is an AI agent for claims handling: it reads the entire file, checks that it is complete, calculates the amount and points to the exact policy clause behind its recommendation. The handler gets a ready-made analysis within minutes.
LiquiAI runs in the background of the system you already use, and sensitive data stays in your local infrastructure. One technology for two roles: the insurer gets a decision recommendation, and the broker gets a review of the insurer’s decision from the client’s point of view.

Automatyzacja likwidacji szkód w liczbach

Wyniki mierzymy na sprawach klienta, których agent wcześniej nie widział.

<1 min zamiast ~15 min na analizę jednej sprawy szkodowej
4-6 tyg. do wyników pilotażu na Twoich danych
0 danych osobowych przekazanych do zewnętrznego modelu AI

For Insurers and Brokers

The insurer assesses whether a claim is valid. The broker defends the client’s interests. LiquiAI works on your side, whichever of the two you represent.
Insurance Companies
  • Claim validity assessment based on policy terms, claims history and business rules.
  • Decision recommendation with the amount and the clause it rests on: payout, rejection or referral for review.
  • Completeness check of the documentation before the case goes to substantive analysis.
  • Consistent decisions regardless of team workload.
  • Scaling without hiring as claim volumes grow.
Insurance Brokers
  • Review of the insurer’s decision against the policy, endorsements, clauses and the master agreement.
  • Rate control: did the handler apply list-price rates instead of the negotiated ones?
  • Review of repair estimates and comparison of photo documentation of the damage.
  • Grounds for appeal based on contract terms, regulations and case law.
  • Control at the notification stage: is the claim being filed under the right policy?

How an AI Agent Works in Claims Handling

– 5 Steps
Book a demo on your own case
1.
Case Intake
LiquiAI automatically pulls the documentation from your claims system.
2.
On-Premise Pseudonymization
A local model, cut off from the internet, reads the documents and replaces sensitive data with identifiers before anything leaves your local infrastructure.
3.
Check Against Policy Terms
The agent compares documents and photos with the policy terms, business rules and claims history. It checks the scope of cover, limits and exclusions.
4.
Ready Recommendation
The handler gets, inside their own system, a completeness assessment, a proposed decision and an amount, backed by the specific clause of the policy terms.
5.
Human Decision
The handler approves, edits or rejects the recommendation. The final word is always theirs, and ambiguous cases are flagged for review right away.

Sensitive Data Never Leaves Your Local Infrastructure

  • Deployment in your local infrastructure. The agent runs on your server, not on ITSharkz servers.
  • Pseudonymization before transmission. Personal data is removed locally before any content reaches a language model. You configure the scope of pseudonymization yourself.
  • EU-hosted models, with no data retention by the model provider.
  • No training on your data. The agent does not learn from your clients’ cases.
  • Full decision trail. Every recommendation comes with a justification ready for control and audit.

One Engine for All Insurance Lines

We tailor the LiquiAI engine to your products, policy terms and business rules. A new line is a matter of configuration, not building from scratch.

Property Claims Handling

Water damage, fires and electronic equipment damage in houses, apartments, businesses and homeowners’ associations.

Travel Claims Handling

Medical expenses abroad, baggage, trip cancellation and delays.

Life Insurance Claims Handling

Benefits under individual and group policies, assessed on the basis of medical records.

Personal Accident Claims Handling

Accident benefits under individual, group and school policies.

Motor Claims Handling

Third-party liability and comprehensive cover, fleets, repair estimates and photo documentation of damage.

GDPR and AI Act Compliance

LiquiAI is designed to help insurers and brokers meet the requirements of the GDPR and the EU Artificial Intelligence Act (AI Act). Compliance comes from the architecture of the solution, not from declarations.
GDPR
  • Data protection by design: pseudonymization happens locally, before data leaves your local infrastructure.
  • Data minimization: only what is needed to assess the case reaches the model.
  • Processing within the European Economic Area: language models are hosted in the EU, with zero data retention.
  • No solely automated decisions: the final decision is always made by the claims handler.
  • You remain the data controller. ITSharkz does not store the personal data of insured persons.
AI Act
  • Human oversight: the agent recommends, the human decides and can correct any recommendation.
  • Transparency: every recommendation comes with a justification and the clause it rests on.
  • Action logging: a full history of analyses and recommendations for audit purposes.
  • Controlled changes: the agent does not learn on its own, and every change is verified on a test set.
  • Audit-supporting documentation: architecture description, risk analysis and pseudonymization test, also for ISO 27001 purposes.

From Workshop to Production in Less Than 10 Weeks

No new system and no overhaul of your process. LiquiAI connects to your claims system via API.

1

Workshop (2 Hours)

Together we review the claims process in the chosen line of insurance and the key business rules the agent will follow.

2

Pilot (4–6 Weeks)

We tailor the agent to your cases, rules and process. The pilot ends with a measurement of effectiveness on an evaluation sample the agent has never seen.

3

Production Rollout (About 4 Weeks)

We launch the agent in your local infrastructure and integrate it with your claims system.

4

Ongoing Development

We supervise the agent's work and regularly calibrate it using measurements from live cases, so its effectiveness keeps improving.

Not sure whether this is the right solution for your company?

Book a free consultation

Book a meeting

The LiquiAI Team

We combine experience in building AI agents with knowledge of the insurance industry: claims processes, policy terms and the day-to-day work of claims handlers. That is why we roll out LiquiAI side by side with claims departments, not next to them.
Kamil Kostrzewski

Kamil Kostrzewski

CTO – solution architecture and security
Jakub Patera

Jakub Patera

Head of AI – agent effectiveness and quality of results
Kacper Kostrzewski

Kacper Kostrzewski

Partnerships with insurers and brokers
Victor Carraz

Victor Carraz

LiquiAI development in the French market

Barthelemy Jaillant

Head of Delivery France
Tomasz Hencel

Tomasz Hencel

Tech Lead

Bohdan Vasylenko

Senior Fullstack Engineer

Michał Drzewiecki

Senior Fullstack Engineer
Learn more about ITSharkz
Meet our team and values
ItSharkz Cube

See LiquiAI on Your Own Case

30 minutes, no strings attached. We’ll show how the agent analyzes the documentation and justifies its recommendation.
Book a demo

Frequently Asked Questions About LiquiAI

What is LiquiAI?

LiquiAI is an AI agent that streamlines the insurance claims handling process, built by ITSharkz, a software company based in Warsaw. It analyzes claim documentation, assesses its completeness and prepares a decision recommendation with the amount and the contract clause it rests on. LiquiAI works for insurers and for brokers.

Which lines of insurance does LiquiAI support?

LiquiAI supports property, motor, life, travel and personal accident claims (individual, group and school). The agent is configured to the products, policy terms and business rules of each client.

Is LiquiAI for insurers or for brokers?

For both. The insurer version assesses the validity of a claim and recommends a decision. The broker version checks the insurer’s decision against the negotiated policy terms and prepares arguments for an appeal in the client’s interest.

Does the AI agent learn from our data on its own?

No. LiquiAI does not learn from your clients’ cases and does not change its behavior without control. Every change, such as new rules or a new document type, goes through controlled calibration and verification on a test set before it reaches production. That keeps results predictable and auditable.

Is LiquiAI GDPR-compliant?

LiquiAI is designed around data protection by design: personal data is pseudonymized locally, in the client’s infrastructure, before it reaches a language model, and processing takes place within the European Economic Area. The client remains the data controller, and the final decision is always made by a person, so there is no solely automated decision within the meaning of Article 22 of the GDPR.

How does LiquiAI meet the requirements of the AI Act?

LiquiAI covers the key requirements of the EU Artificial Intelligence Act: human oversight of every decision, transparent recommendations, action logging and controlled changes to the agent’s behavior. ITSharkz prepares architecture documentation and a risk analysis that help the client assess compliance.

Is our clients’ personal data sent to an external AI model?

No. LiquiAI removes personal data locally, in the client’s infrastructure, before any content reaches a language model. The model responsible for pseudonymization has no network access. The language models are hosted in the European Union.

Who decides on the claim payout?

Always a person. LiquiAI prepares a recommendation with a justification, and the claims handler approves, edits or rejects it. Ambiguous cases are flagged for review by the agent itself.

How is LiquiAI different from an insurer’s in-house AI tools?

LiquiAI is a ready-made agent proven on real claims, so a pilot on the client’s data takes 4 weeks instead of a difficult in-house project that can run for months or even years. It can also run alongside existing tools, covering lines or process stages the in-house solution does not.

Will the pilot results carry over to production?

LiquiAI is calibrated on one sample of cases, and its effectiveness is measured on a completely separate one. The pilot result therefore shows how the agent handles cases it has never seen.

Does deploying LiquiAI require changing our claims system?

No. LiquiAI works in the background of your existing claims system and connects to it via API. The handler sees the result in the tool they already use.

Will LiquiAI pass an ISO 27001 audit?

LiquiAI is designed with audits in mind: it runs in the client’s infrastructure, pseudonymizes data locally and documents the justification for every recommendation. ITSharkz supports the client with the risk analysis and pseudonymization tests the auditor requires.

How long does deployment take?

A pilot on the client’s data takes 4–6 weeks from receiving the cases, and the production rollout takes about 4 weeks after the pilot is accepted. The whole process from workshop to production usually takes 10–12 weeks.

How can we check whether LiquiAI will work on our cases?

The fastest way is a demo: in 30 minutes we show the agent on a real case from your line of insurance. The next step is a pilot on a sample of your cases, with a success threshold agreed before the start.