Devoteam helped Business Post, Ireland's only privately-owned national media company, modernise more than 55,000 legacy property addresses to Ireland's post-2015 Eircode system using AWS Bedrock (powered by Anthropic Claude 3.5 Sonnet), a custom geo-matching algorithm, and AI-driven satellite-imagery analysis — achieving an 85% match rate against a 70% target across 15,669 planning applications.
"The technical team at the Business Post Group has found real value in collaborating with Devoteam. Their structured approach brings a diverse range of strengths to the table, enabling us to address challenges and capitalize on opportunities with impressive speed and tangible results," said Craig Tait, Group Technical & Data Director at Business Post Group.
The modernisation was driven by an address-system mismatch. Business Post had acquired a property, business, and legal dataset whose records used pre-2015 Irish addressing conventions, making the data hard to search, link, and reuse. Matching the records by hand would have been time-consuming and potentially error-prone. Resolving the problem would let Business Post apply the data internally for journalism and license it to real estate, finance, legal services, and market research.
Devoteam delivered the work as a three-phase POC covering County Carlow. Phase 1 set up a sandbox with Claude 3.5 Sonnet and enriched the dataset with Irish Government open data. Phase 2 built a two-stage geo-matching algorithm: first matching by latitude/longitude and planning polygons, then using AI-driven satellite imagery to refine ambiguous cases. Phase 3 covered validation against known data and live demos.
Technically, the solution combines four AWS services: AWS Bedrock (Claude 3.5 Sonnet) for understanding historical address formats, AWS Location for precise geocoding, AWS Lambda and Notebooks for serverless processing, and AWS S3 for storage. The system was designed to scale from the County Carlow POC to nationwide use.
At operational scale, the POC processed 55,000+ locations across 15,669 planning applications, with matches breaking down into 55% exact, 28% AI (resolved through satellite imagery — for cases like shopping centres), 2% close (nearby locations for large properties like farmland), and 15% no match. Detailed testing on 1,033 applications containing 3,783 locations validated handling of complex scenarios like multiple buildings per application and boundary changes.
Looking ahead, Business Post is using the POC results as the foundation for a broader digital transformation. Business Post is planning a migration across its entire dataset through the AWS Migration Acceleration Program (MAP) — enabling faster property research for journalists, new data-driven products for subscribers, potential licensing across real estate, legal, and finance, and enriched digital content. The company is also aiming to grow its digital-only news business from 10,000 to 40,000 subscribers as part of the same transformation.