The Bank of England has leveraged artificial intelligence to analyze firms’ pricing decisions using text-based business intelligence. Researchers Marcus Buckmann, Galina Potjagailo, and Shobhit Rawat employed a large language model (LLM) to quantify pricing narratives from company visit reports. The LLM derived scores for output-price changes and decomposed them into factors like direct and indirect input costs, labor costs, and demand.
The study analyzed 66,820 reports from January 2005 to September 2026, with a focus on a random one-third sample up to 2025 and the full sample for 2026. The LLM tracked realized output prices closely with CPI inflation, showing greater month-to-month volatility. It also identified key determinants such as direct energy-related costs, labor costs, and demand, providing a nuanced view of firms’ pricing decisions over time.
During the global financial crisis, weak demand dominated price pressures, while the recent inflation surge highlighted labor costs and strategic pricing. The Middle East conflict has introduced new cost pressures, but these remain less broad and persistent than those observed in 2022. The LLM’s decomposition of pricing determinants offers valuable insights into the mechanisms driving inflation.