Accelerating Urban Planning with a Secure AI-Powered Chatbot

A major Saudi Arabian giga-project sought to streamline access to complex urban planning information. Oivan was selected to develop and implement a sophisticated, secure AI-powered chatbot to meet this need.

Case Study Highlights

Client

Saudi Giga Project

Year(s)

2024

Service

AI & Emerging Technologies Development

Introduction

As part of its vision to be a technology leader and enabler of Saudi Vision 2030, a major Saudi Arabian giga-project sought to streamline access to complex urban planning information. Its Planning Department required an intelligent solution that would provide quick, accurate answers to stakeholder queries while maintaining the highest standards of data privacy. Oivan, a long-established digital transformation consultancy operating in the Kingdom, was selected to develop and implement a sophisticated, secure AI-powered chatbot to meet this need.

Challenge

The client’s Planning Department manages a vast repository of technical documentation – including more than 50 key planning documents and a frequently updated set of FAQs. Delivering timely, accurate, and consistent responses to internal and external stakeholders presented three main challenges:

Intelligent: Understanding natural language queries and providing contextually relevant, technically accurate responses.
Secure: Ensuring full data sovereignty by hosting all systems within the client’s private data center – essential for compliance with national and organizational security standards.
Efficient: Enabling rapid access to complex planning data without manual effort from subject-matter experts.

Solution

Oivan designed and implemented a hybrid chatbot architecture that combines a rules-based system with an LLM-powered retrieval engine, offering both precision and conversational depth.

The system architecture includes:

Dual-Source Information Retrieval: For common queries, the chatbot first consults a predefined set of FAQs cultivated by the Planning Department. If the answer is not found, the query is passed to an LLM agent.
Advanced AI Integration: The LLM agent utilizes Mistral 7B, an open-source model, combined with Retrieval-Augmented Generation (RAG). This RAG system searches a vector database containing over 50 indexed planning documents to find the most relevant information and generate a comprehensive, context-aware response.
Secure, On-Premise Hosting: The system runs fully on-premise within the client’s data center, powered by NVIDIA’s GPUs – guaranteeing compliance with strict data protection protocols.
Open-Source Framework: Built using the Rasa conversational framework for intent recognition, response generation, and seamless dialogue orchestration.

Outcome

The AI-powered chatbot, integrated into the client’s internal web portal, provides an intelligent and efficient assistant for urban planning knowledge:

Enhanced Accessibility: Stakeholders can now receive instant, natural-language answers to complex questions about land use, compliance, and planning policies directly through the portal.
Improved Efficiency: The system automates responses to a wide range of queries, freeing up the Planning Department’s staff to focus on more strategic tasks.
Guaranteed Data Security: By deploying the solution entirely on-premise, Oivan ensured that all sensitive planning data remains within the client’s secure infrastructure, a critical requirement for the project.
Scalable Foundation: The modular architecture allows for future enhancements, such as expanding the knowledge base with documents from other departments or upgrading the LLM to a more advanced version.

This project showcases Oivan’s ability to deliver cutting-edge, secure, and customized AI solutions that address the unique challenges of large-scale, visionary enterprises with strict data protection and security requirements

FAQ

Q1: Can similar AI systems be used in other government or private sectors?

A: Yes. Oivan’s AI architecture can be adapted for ministries, utilities, and real estate developers seeking to deploy private, domain-specific AI assistants under local data governance regulations.

Q2: Why use open-source LLMs like Mistral 7B?

A: Open-source models allow full control over data, performance, and security — crucial for clients who need on-premise deployment or want to avoid exposure to external APIs.

Q3: How does Oivan ensure data security in critical AI deployments?

A: Architecture can be designed so that all processing occurs within isolated environments managed by the client, with no internet dependencies, ensuring data sovereignty and full compliance with Saudi data privacy and cybersecurity frameworks.

Let’s Work Together

Whether you want to discuss the latest developments in digital service innovation, or have an idea for a service for your organization, we are always happy to talk.