Descriptions:
Jess Grogan-Avignon and Jack Wang from Accenture present a practitioner-focused talk at the AI Engineer conference, drawing on firsthand experience deploying agentic AI inside some of the world’s largest enterprises — telecoms, utilities, healthcare systems, and government agencies. Their central argument: most enterprise agentic projects are set up to fail not because of model capability gaps, but because organizational scaffolding designed for human-paced work cannot handle machine-speed delivery.
The talk is structured around five enterprise tensions the Accenture team has identified in the field. These include the mismatch between AI delivery speed and enterprise approval infrastructure (one project took two weeks to build but twelve months to reach production), the need for finance functions to adopt a VC portfolio mindset rather than demanding predictable ROI before greenlighting agentic bets, and the category error of scoping non-deterministic agent behavior like a fixed software feature build. They cite GitHub’s data — 1 billion commits reported in 2025, accelerating to a projected 14 billion by end of 2026 at current weekly rates — as evidence that code supply is exploding while deployment infrastructure has barely changed. Accenture’s own research shows only 12% of companies reach what they call “AI achiever” status.
Case studies include Walmart building a generative social-media trend scanner to compete with Shein and Temu, and JP Morgan productizing an internal AI productivity tool into a net-new revenue stream. The talk is aimed at AI engineers and enterprise architects navigating the gap between pilot success and production scale.
📺 Source: AI Engineer · Published May 28, 2026
🏷️ Format: Deep Dive
