This AI Called 6,500 Pubs to Find the Cheapest Guinness

Jan de Vries · 2026-11-13

In a recent article, we looked at iBeer, the viral app that let you pretend to drink beer from your phone. Another tech experiment took a similarly playful idea in a more practical direction, asking a simple question: "How much is a pint of Guinness?" The answer led to Guinndex.ai, an AI-powered, crowdsourced index tracking Guinness prices across Ireland. Behind the project is Matt Cortland, an AI engineer and former pub owner who wondered whether he was overpaying for his pint. The concept wasn't entirely new. Back in 2017, three university students began developing the original Guindex, a community-driven initiative to track Guinness prices, initially focusing on Dublin. The project expanded across the Republic of Ireland, using OpenStreetMap to identify pubs and inviting the public to contribute prices through a website launched in June 2018. Nearly a decade later, Cortland independently revisited the concept with a different approach. Instead of relying primarily on crowdsourced information, he used AI to automate much of the data collection process. ### How does it work? At its core, the Guinndex brings together an existing stack of technologies into a relatively straightforward data collection pipeline: - Pub discovery: Google's Places API identifies and maps more than 6,500 pubs across Ireland's 32 counties. - AI-powered calls: An agent named Rachel, built using ElevenLabs' conversational AI technology and supported by Twilio's telephony infrastructure, contacts pubs to ask about their Guinness prices. - Data extraction: Anthropic's Claude processes call transcripts to identify and record the quoted prices. - Website and database: Next.js and Supabase help power the searchable index, while Claude Code supports development. - Community contributions: Pub owners and customers can submit new prices and correct existing listings, helping keep the data relevant. Having put the stack together, it still needed some refining in practice. Early versions of Rachel's script included additional questions and repeated prices for confirmation. However, these longer conversations proved less effective. The solution was to simplify. Rachel would ask the price, thank whoever answered and end the call, leaving Claude to extract the information from the transcript afterwards. It's a small but telling design decision; rather than adding more features, Cortland made the system more effective by removing unnecessary steps. ### So, what are the results? The Guinndex now lists more than 6,500 pubs and thousands of verified prices. At the time of writing, its published all-Ireland average stood at €6.06 (about $6.60), with €5.50 (around $6.00) remaining the most commonly recorded price. This is bookended by the highest all-Ireland price of €11.00 (roughly $12.00) and cheapest at €3.00 (about $3.30). Geographic differences are substantial. Unsurprisingly, Dublin is among the most expensive locations, with an average of €6.93 (around $7.55) and three pubs charging €11.00 (about $12.00). However, the index also identifies pubs in the city and its suburbs where a pint costs less than €5.00 (roughly $5.45). Pubs in smaller towns and rural areas often offer more affordable prices with averages sometimes dipping just below the national level. Of course, the data has limitations. Some pubs don't or didn't answer calls, prices can change, and community submissions may contain errors. Unlike an official statistical survey, the Guinndex depends on automated collection and user contributions. Nevertheless, its value lies in making information searchable, accessible and useful to the people interested in it. > "The goal isn't to be perfect, it's to be useful. And if it saves someone a few euros on their next pint, that's a win." – Matt Cortland The approach also raises questions about transparency when AI agents interact with businesses. While such calls can make data collection more efficient, their wider adoption will depend partly on ensuring that people know when they are speaking to an automated system. As such, The Guinndex subsequently published a notice stating that any future calls would begin with a clear disclosure that the caller is an AI. ### What this means for AI and data collection The Guinndex is a neat example of how AI can turn a simple question into a valuable dataset. It also shows the importance of thoughtful design—sometimes less is more. As AI agents become more common, transparency will be key to maintaining trust. For now, the Guinndex offers a glimpse into a future where AI helps us find answers to everyday questions, one pint at a time.