PZU i360: an AI-orchestrated broker inquiry intake Qorus-NTT DATA Innovation in Insurance Awards 2026

Concept & Objectives
PZU i360 rewrites the first mile of corporate insurance. The AI‑driven workflow ingests broker emails and diverse attachments, classifies inquiries by product line, and extracts quote-critical fields straight to the policy system, so sales teams start from insight, not inboxes. The objective: a step‑change in speed, accuracy and consistency that makes straight‑through processing the default, not the exception. The AI-powered workflow is designed to shift staff from manual parsing to exception-based work and higher-value client tasks. This initiative delivers measurable operational acceleration (STP and cycle‑time gains), governed AI at scale, and an embedded frontline capability for data quality and AI orchestration that drives continuous improvement.

Reasons Behind
Sales teams in the corporate insurance line handle annually ~40–50k broker inquiries in unstructured formats (email bodies, PDFs, images, Office files). Manual parsing created bottlenecks, cycle‑time drag and error risk. We set out to replace repetitive, low‑value work with governed AI that delivers clean, structured, ready‑to‑use data for sales and underwriting — at enterprise scale.

State of Competition

In Poland’s corporate insurance market, handling broker inquiries is still predominantly manual, with advisors parsing emails and mixed‑format attachments— it is a sector‑wide challenge, not unique to PZU. As a result, AI‑enabled straight‑through intake is rare across domestic competitors today, and PZU’ i360 solution stands out as an innovative, market‑leading initiative: an end‑to‑end, AI‑orchestrated intake that converts unstructured broker emails and attachments into structured, quote‑ready data.

Sources of Inspiration
The program advances PZU’s broader digitalization agenda: automate knowledge work, raise data quality at the source, and reuse extracted information across the value chain—from intake to underwriting preparation and beyond—all under the banner of responsible, enterprise‑grade AI.

Departments Involved
A cross‑functional coalition underpinned the pilot and scale design:

• Corporate Client Department (business ownership and SME validation),

• Innovation & AI Team (solution ownership and project management),

• IT/Solutions & Cybersecurity (integration, controls),

• AI governance (responsible AI & monitoring),

• Procurement (negotiations).

Main Results So Far

• Scale & robustness validated: The workflow was proven on a representative real‑world set of broker emails and mixed‑format attachments, confirming reliability across PDFs, images, Office files and ZIP files

• Speed to action: Intake now converts inbox content into quote‑ready data in minutes per case, demonstrating meaningful cycle‑time compression for front‑line teams.

• Live reliability with continuous improvement: Live testing confirmed high data quality, while targeted prompt‑engineering materially improved the most challenging fields, establishing a durable learning loop

• Enterprise‑ready & governed: The end‑to‑end flow (mailbox → AI extraction → RPA → policy system) operates with admin/reporting controls and aligns with security and regulatory expectations, including a positive AI‑Act‑related assessment.

• From pilot to business as usual: The solution is now transitioning from pilot to business‑as‑usual, with production rollout in progress across corporate sales intake.