The Language of AI Deployment

Abstract

Between January 2022 and April 2026, corporate leaders publicly described their use of artificial intelligence thousands of times. This study examines 155 verified public statements from nine major organisations: Microsoft, IBM, Salesforce, Shopify, Duolingo, Cisco, Meta, Amazon, and JPMorgan Chase. Drawing on press releases, earnings calls, shareholder letters, regulatory filings, and executive interviews, each statement was verified verbatim and coded against five descriptive registers: Descriptor, Expansion, Substitution, Walk-back, and Operational. The coded record reveals that expansion language dominates corporate messaging, substitution claims concentrated late in the window, and public qualifications arrived on timetables ranging from weeks to never. Rather than moving through fixed phases, these five categories function as fluid registers that coexist across speakers and channels. Crucially, while organisations frequently soften or qualify their public language following pushback, internal AI deployment continues without interruption.