PZU

As Poland's largest insurer and the biggest financial institution in Central and Eastern Europe, PZU handles an enormous volume of auto claims annually. Under its traditional process, damage assessments required dispatching or coordinating human adjusters to evaluate vehicle damage after an accident — a process that stretched claim resolution to days. For policyholders already stressed by an accident, this delay created friction at exactly the wrong moment. At scale, slow cycle times also drive up claims handling costs and erode customer satisfaction scores. The core challenge was compressing the most time-sensitive step — initial damage assessment at first notice of loss — without sacrificing accuracy.

PZU deployed Tractable's AI Estimating solution at the first notice of loss (FNOL) stage, integrating computer vision directly into the claim intake workflow. When a policyholder reports an accident, they submit smartphone photos of the damaged vehicle through PZU's existing claims channel. Tractable's deep learning model — trained on millions of vehicle damage photographs — analyzes the images, identifies affected parts, recommends repair operations, and calculates repair costs, typically without requiring human intervention. The FNOL deployment built on a partnership that began in 2017, during which Tractable's AI was already running in a back-office capacity to audit bodyshop repair quality. Expanding to customer-facing FNOL automation represented a deliberate, staged progression from internal quality control to front-line claim handling.

The impact was immediate and measurable. Claims that previously took days to resolve can now be settled in minutes, with policyholders receiving damage assessments and cash settlement options almost immediately after submitting photos. Key outcomes include:

• Tens of thousands of Polish auto claims processed annually through the AI pipeline
• Several hundred thousand total claims handled using Tractable-powered tools prior to the FNOL expansion, representing several billion PLN in claim value
• PZU became the first Polish insurer to offer AI-driven FNOL assessment, and ranks among the largest AI deployments in the global auto repair ecosystem
• Claim handlers can communicate results and offer settlements almost instantly, giving policyholders agency to accept cash or pursue repair

• FNOL is the highest-leverage integration point — applying AI at first notice of loss compresses the entire claims cycle, not just one back-office step.
• Staged rollouts reduce deployment risk — PZU's seven-year progression from quality audits to customer-facing automation let the organization validate the AI before entrusting it with policyholder-facing decisions.
• Instant cash settlement changes the customer experience — offering an immediate payment option eliminates the need for a full process redesign while dramatically cutting cycle time.
• Deep learning models require scale to generalize — a system trained on millions of damage images can handle the variability of real-world accidents reliably, which is a prerequisite before deploying at FNOL.