How Arcadis built a repeatable AI communication model for public infrastructure projects with Chatbase

At a glance

Arcadis is a global engineering and design consultancy with deep involvement in major public infrastructure projects. Peter Schuch and his team manage projects with significant public impact, where transparent communication with citizens is part of the brief. One such project is W27 on the Lower Rhine, a fairway clearance program that runs through Düsseldorf Harbor, with all works carried out from the waterway. For projects of this scale, communication is part of the deliverable. Arcadis needed a way to make official, curated project information accessible to citizens without burying them in engineering documentation.

A complex infrastructure project, and a public that needed answers

Major infrastructure projects come with a built-in communication challenge. Citizens want to understand what is being built, why, where, and how it affects them. Project documentation exists, but it is technical, fragmented, and not designed for the general public. For W27, which involves removing a Mittelgrund and adding an open Längswerk at Teilabschnitt 2 Lausward to keep the fairway to Düsseldorf Harbor clear, the team needed a channel that could run 24/7 with no waiting times, answer questions in plain language, pull from official curated WSA Rhein documentation, and guide citizens to the right contacts and participation opportunities. PDFs, project pages, and info hotlines do not scale and do not invite participation. Peter's team needed an interactive layer that met the compliance bar required for public-sector communication.

KIra Rhein, an AI agent built for project transparency

Arcadis built KIra Rhein, an AI agent powered by Chatbase, trained on official, curated project information from the WSA Rhein. Citizens can ask the agent questions about the project and get answers, definitions, and links to deeper resources. The agent also guides citizens to the right contacts and participation opportunities, turning a one-way information page into an interactive resource. Because the project serves a German governmental organization, Arcadis implemented Mistral, a European model, to keep the deployment aligned with public-sector data sovereignty expectations. The build was straightforward. The compliance review, given the public-sector context, took several months. The result is an AI agent that meets the bar for official communication on a public infrastructure project, while still feeling accessible and easy to use for the general public.

The results

• KIra Rhein is live on the W27 project, giving citizens 24/7 access to official project information in conversational form.
• The underlying information is not new, it exists in WSA Rhein documentation. The access is. Instead of digging through technical documents, citizens can ask a question and get a direct answer with links to source material and the right people to contact.
• The agent meets the compliance bar required for public-sector communication while delivering the citizen experience public projects increasingly need.
• Citizens use it more and more as awareness of how to get information from AI grows.