Business Post: Modernising Planning Data with GenAI

About Business Post

The Business Post stands as a cornerstone of Irish media, distinguished as the country’s only privately-owned national media company. With iconic brands like Irish Tatler and Food & Wine in its portfolio, the organisation has built its reputation on trusted, respected, and independent journalism through both organic growth and strategic acquisitions.

The company is embarking on an ambitious digital transformation journey, aiming to expand its digital-only news business from 10,000 to 40,000 subscribers. This modernisation strategy focuses on leveraging data and AI to drive digital growth and global expansion, accelerated by the evolving demands of on-demand news consumption and the opportunities presented by emerging technologies.

The challenge

Business Post recently acquired a comprehensive dataset spanning property, business, and legal domains. While initially acquired to support their journalism, the organisation recognised these records could have broader applications across various industries.

To explore the potential of this dataset, Business Post decided to focus first on the property information, which presented a clear challenge and use case. The property records included an extensive collection of ‘Site Plans’ with addresses that predated a significant change in Ireland’s addressing system.

In 2015, Ireland introduced the Eircode system, assigning a unique identifier to every address in the country. This new standard created a need to update and align existing address data with the Eircode format to maintain relevance and usability.

Matching the pre-existing addresses in Business Post’s records to the new Eircode system presented several challenges:

• Legacy addressing formats made it difficult to track building modifications and legal property information
• Efficient data searching and linking was hindered by outdated address formats
• Manual matching would be time-consuming and potentially error-prone
• The full value of the dataset was difficult to access due to the outdated reference system

 Business Post recognised that resolving these challenges could not only enhance their own use of the data but also potentially create opportunities in sectors such as real estate, finance, legal services, and market research.

Methodology: The Development Journey

Our approach combined careful preparation with innovative development, ensuring each phase built upon the last to create a robust, scalable solution.

Phase 1: Data & environment initiation

We began by laying strong foundations. Setting up a sandbox environment with the Anthropic Claude 3.5 Sonnet model gave us a powerful AI engine to work with. Alongside this, we undertook extensive data refinement and normalisation, enriching our dataset with Irish Government open data to improve matching accuracy.

Phase 2: Implementation & development

With clean data and a stable environment in place, we focused on building the core solution. Our team developed a sophisticated geo-matching algorithm for precise location identification, supported by a robust data processing pipeline.

The matching process worked in two stages:

• First, we utilised latitude and longitude coordinates from both property records and planning application polygons, implementing an interception method to identify potential matches.
• Then, for cases where multiple locations were identified, we employed AI-driven analysis of satellite imagery and contextual planning data to refine the results.

This dual approach allowed us to accurately match potential Eircodes by considering spatial relationships, satellite imagery, and contextual information within the planning applications

Phase 3: Validation & demo

Rigorous testing was crucial for a solution handling sensitive property data. We conducted comprehensive system testing, validated results against known data points, and refined the matching logic based on real-world scenarios. Live demonstrations to stakeholders confirmed the system’s effectiveness, while thorough documentation ensured smooth knowledge transfer.

AWS utilisation

Our solution architecture leveraged four key AWS services, each chosen to address specific aspects of the challenge:

• AWS Lambda / Notebooks: Provided serverless processing power for data analysis and transformation, enabling cost-effective scaling based on demand. Notebooks offered a flexible environment for developing and refining our matching algorithms.
• AWS Bedrock: Through the Anthropic Claude 3.5 Sonnet Model, we gained sophisticated AI capabilities for understanding complex address formats and contextual information. This foundation model proved particularly effective at interpreting historical addressing conventions.
• AWS Location: Delivered precise geo-matching capabilities essential for validating address matches. Its robust geocoding services helped resolve ambiguous locations and confirm spatial relationships between historical and modern addresses.
• AWS S3: Offered secure, scalable storage for both raw and processed data. The service’s high durability and easy integration with other AWS services made it ideal for managing our growing dataset of enriched property information.

Results overview

Our solution successfully transformed historical records into a modern, searchable resource for the proof of concept area of County Carlow. The system processed over 55,000 locations across 15,669 planning applications, achieving an outstanding 85% match rate between historic Site Plans and modern Eircodes, paving the way for potential nationwide implementation.

This success broke down into distinct matching categories, each serving specific business needs:

• Exact match (55%): Perfect alignment with modern addresses, enabling immediate property identification for houses and apartment blocks
• AI match (28%): Complex cases resolved through AI analysis of application descriptions and satellite data, particularly valuable for developments like shopping centres
• Close match (2%): Nearby location matches, useful for large properties like farmland where exact boundaries may have changed
• No match (15%):  Cases with no address data on planning application or manual review

To validate our approach, we conducted detailed testing on a focused batch of 1,033 applications containing 3,783 associated locations. This targeted analysis proved the system’s capability to handle complex scenarios, including:

• Multiple buildings sharing a single planning application
• Properties with boundary changes over time
• Locations with significant historical development
• Areas with substantial address format changes

The high success rate and robust handling of complex cases demonstrates the solution’s readiness for full-scale deployment across Business Post’s entire dataset.

Strategic outcome

The success of this proof of concept has laid the groundwork for Business Post’s broader digital transformation initiative.  This solution provides a crucial foundation for enriched content delivery and new data services.

Building on these results, Business Post is now actively planning a comprehensive migration strategy through the AWS Migration Acceleration Program (MAP). This strategic phase is designed to scale the solution across their entire dataset, enabling:

• Faster, more accurate property research for journalists
• New data-driven products for subscribers
• Potential licensing opportunities across real estate, legal, and financial sectors
• Enhanced digital content enriched with historical context

The high match rates and sophisticated AI capabilities demonstrated in the POC validate Business Post’s vision of transforming from a traditional media company into a data-driven, digital-first organisation.

Why Devoteam?

Devoteam is an AWS Premier Partner specialising in cloud migration, modernisation, and Generative AI. Our expertise in the AWS Well-Architected Framework ensures every solution we deliver follows established best practices.

What sets us apart:

• AWS Premier Partner status: Recognition of our consistently high standards in AWS solution delivery
• GenAI specialists: Practical experience implementing AWS Bedrock and other AI solutions at scale
• Migration expertise: Deep experience with AWS Migration Acceleration Program (MAP)
• Architecture excellence: Solutions built on Well-Architected Framework principles
• Knowledge transfer focus: We ensure your team is equipped for long-term success