Token prices have collapsed, yet the AI bill keeps going up. And the data your agents need is still stuck in documents no machine was meant to read. ​This Future Form covers the two unglamorous problems that decide whether AI holds up in production: what it costs to run, and whether you can trust what comes out. Expect a closer look at the tools, approaches and trade-offs involved, with ideas you can take back to your own projects. ​Hosted by BlackRock. Pizza and drinks included. ​**Joe Petterson, Partner, QuantumBlack, AI by McKinsey** ​*Your agents have a spending problem* ​Why most teams burn through their AI budget by mid-year, and the engineering levers that cut inference cost by about half. Live demos: 16 models head to head on one task, quality against cost, and the point where self-hosting actually pays back. ​**Thomas Tosstorff, Ines Blanco and Ed Carey, BlackRock** ​*Nobody's documents are clean* ​Every team ends up building its own document extraction pipeline. BlackRock decided to build one set of shared capabilities instead. Thomas, Ines and Ed share what they learned turning unstructured documents into structured, actionable data, and where AI actually earns its place in the process. Join us for pizza, drinks, and practical insights of where AI can genuinely accelerate delivery, and what needs to change to make that happen. ​**If you’re working hands-on with AI tools or responsible for how software gets delivered, this session is for you.**

Description supplied by the event organiser.